shu/little-librarian
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

initial commit

Browse source
This commit is contained in:
Sebastian Hugentobler 2025-07-01 08:55:08 +02:00
commit 0bd97d0ed3
Signed by: shu
SSH key fingerprint: SHA256:ppcx6MlixdNZd5EUM1nkHOKoyQYoJwzuQKXM6J/t66M
25 changed files with 6581 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

1
.envrc Normal file
View file

@ -0,0 +1 @@
use flake

8
.gitignore vendored Normal file
View file

@ -0,0 +1,8 @@
.direnv/
target/
rustc
.DS_Store
model.onnx
tokenizer.json
reranking.onnx
db_url

661
COPYING Normal file
View file

@ -0,0 +1,661 @@
GNU AFFERO GENERAL PUBLIC LICENSE
Version 3, 19 November 2007
Copyright (C) 2007 Free Software Foundation, Inc. <https://fsf.org/>
Everyone is permitted to copy and distribute verbatim copies
of this license document, but changing it is not allowed.
Preamble
The GNU Affero General Public License is a free, copyleft license for
software and other kinds of works, specifically designed to ensure
cooperation with the community in the case of network server software.
The licenses for most software and other practical works are designed
to take away your freedom to share and change the works. By contrast,
our General Public Licenses are intended to guarantee your freedom to
share and change all versions of a program--to make sure it remains free
software for all its users.
When we speak of free software, we are referring to freedom, not
price. Our General Public Licenses are designed to make sure that you
have the freedom to distribute copies of free software (and charge for
them if you wish), that you receive source code or can get it if you
want it, that you can change the software or use pieces of it in new
free programs, and that you know you can do these things.
Developers that use our General Public Licenses protect your rights
with two steps: (1) assert copyright on the software, and (2) offer
you this License which gives you legal permission to copy, distribute
and/or modify the software.
A secondary benefit of defending all users' freedom is that
improvements made in alternate versions of the program, if they
receive widespread use, become available for other developers to
incorporate. Many developers of free software are heartened and
encouraged by the resulting cooperation. However, in the case of
software used on network servers, this result may fail to come about.
The GNU General Public License permits making a modified version and
letting the public access it on a server without ever releasing its
source code to the public.
The GNU Affero General Public License is designed specifically to
ensure that, in such cases, the modified source code becomes available
to the community. It requires the operator of a network server to
provide the source code of the modified version running there to the
users of that server. Therefore, public use of a modified version, on
a publicly accessible server, gives the public access to the source
code of the modified version.
An older license, called the Affero General Public License and
published by Affero, was designed to accomplish similar goals. This is
a different license, not a version of the Affero GPL, but Affero has
released a new version of the Affero GPL which permits relicensing under
this license.
The precise terms and conditions for copying, distribution and
modification follow.
TERMS AND CONDITIONS
0. Definitions.
"This License" refers to version 3 of the GNU Affero General Public License.
"Copyright" also means copyright-like laws that apply to other kinds of
works, such as semiconductor masks.
"The Program" refers to any copyrightable work licensed under this
License. Each licensee is addressed as "you". "Licensees" and
"recipients" may be individuals or organizations.
To "modify" a work means to copy from or adapt all or part of the work
in a fashion requiring copyright permission, other than the making of an
exact copy. The resulting work is called a "modified version" of the
earlier work or a work "based on" the earlier work.
A "covered work" means either the unmodified Program or a work based
on the Program.
To "propagate" a work means to do anything with it that, without
permission, would make you directly or secondarily liable for
infringement under applicable copyright law, except executing it on a
computer or modifying a private copy. Propagation includes copying,
distribution (with or without modification), making available to the
public, and in some countries other activities as well.
To "convey" a work means any kind of propagation that enables other
parties to make or receive copies. Mere interaction with a user through
a computer network, with no transfer of a copy, is not conveying.
An interactive user interface displays "Appropriate Legal Notices"
to the extent that it includes a convenient and prominently visible
feature that (1) displays an appropriate copyright notice, and (2)
tells the user that there is no warranty for the work (except to the
extent that warranties are provided), that licensees may convey the
work under this License, and how to view a copy of this License. If
the interface presents a list of user commands or options, such as a
menu, a prominent item in the list meets this criterion.
1. Source Code.
The "source code" for a work means the preferred form of the work
for making modifications to it. "Object code" means any non-source
form of a work.
A "Standard Interface" means an interface that either is an official
standard defined by a recognized standards body, or, in the case of
interfaces specified for a particular programming language, one that
is widely used among developers working in that language.
The "System Libraries" of an executable work include anything, other
than the work as a whole, that (a) is included in the normal form of
packaging a Major Component, but which is not part of that Major
Component, and (b) serves only to enable use of the work with that
Major Component, or to implement a Standard Interface for which an
implementation is available to the public in source code form. A
"Major Component", in this context, means a major essential component
(kernel, window system, and so on) of the specific operating system
(if any) on which the executable work runs, or a compiler used to
produce the work, or an object code interpreter used to run it.
The "Corresponding Source" for a work in object code form means all
the source code needed to generate, install, and (for an executable
work) run the object code and to modify the work, including scripts to
control those activities. However, it does not include the work's
System Libraries, or general-purpose tools or generally available free
programs which are used unmodified in performing those activities but
which are not part of the work. For example, Corresponding Source
includes interface definition files associated with source files for
the work, and the source code for shared libraries and dynamically
linked subprograms that the work is specifically designed to require,
such as by intimate data communication or control flow between those
subprograms and other parts of the work.
The Corresponding Source need not include anything that users
can regenerate automatically from other parts of the Corresponding
Source.
The Corresponding Source for a work in source code form is that
same work.
2. Basic Permissions.
All rights granted under this License are granted for the term of
copyright on the Program, and are irrevocable provided the stated
conditions are met. This License explicitly affirms your unlimited
permission to run the unmodified Program. The output from running a
covered work is covered by this License only if the output, given its
content, constitutes a covered work. This License acknowledges your
rights of fair use or other equivalent, as provided by copyright law.
You may make, run and propagate covered works that you do not
convey, without conditions so long as your license otherwise remains
in force. You may convey covered works to others for the sole purpose
of having them make modifications exclusively for you, or provide you
with facilities for running those works, provided that you comply with
the terms of this License in conveying all material for which you do
not control copyright. Those thus making or running the covered works
for you must do so exclusively on your behalf, under your direction
and control, on terms that prohibit them from making any copies of
your copyrighted material outside their relationship with you.
Conveying under any other circumstances is permitted solely under
the conditions stated below. Sublicensing is not allowed; section 10
makes it unnecessary.
3. Protecting Users' Legal Rights From Anti-Circumvention Law.
No covered work shall be deemed part of an effective technological
measure under any applicable law fulfilling obligations under article
11 of the WIPO copyright treaty adopted on 20 December 1996, or
similar laws prohibiting or restricting circumvention of such
measures.
When you convey a covered work, you waive any legal power to forbid
circumvention of technological measures to the extent such circumvention
is effected by exercising rights under this License with respect to
the covered work, and you disclaim any intention to limit operation or
modification of the work as a means of enforcing, against the work's
users, your or third parties' legal rights to forbid circumvention of
technological measures.
4. Conveying Verbatim Copies.
You may convey verbatim copies of the Program's source code as you
receive it, in any medium, provided that you conspicuously and
appropriately publish on each copy an appropriate copyright notice;
keep intact all notices stating that this License and any
non-permissive terms added in accord with section 7 apply to the code;
keep intact all notices of the absence of any warranty; and give all
recipients a copy of this License along with the Program.
You may charge any price or no price for each copy that you convey,
and you may offer support or warranty protection for a fee.
5. Conveying Modified Source Versions.
You may convey a work based on the Program, or the modifications to
produce it from the Program, in the form of source code under the
terms of section 4, provided that you also meet all of these conditions:
a) The work must carry prominent notices stating that you modified
it, and giving a relevant date.
b) The work must carry prominent notices stating that it is
released under this License and any conditions added under section
7. This requirement modifies the requirement in section 4 to
"keep intact all notices".
c) You must license the entire work, as a whole, under this
License to anyone who comes into possession of a copy. This
License will therefore apply, along with any applicable section 7
additional terms, to the whole of the work, and all its parts,
regardless of how they are packaged. This License gives no
permission to license the work in any other way, but it does not
invalidate such permission if you have separately received it.
d) If the work has interactive user interfaces, each must display
Appropriate Legal Notices; however, if the Program has interactive
interfaces that do not display Appropriate Legal Notices, your
work need not make them do so.
A compilation of a covered work with other separate and independent
works, which are not by their nature extensions of the covered work,
and which are not combined with it such as to form a larger program,
in or on a volume of a storage or distribution medium, is called an
"aggregate" if the compilation and its resulting copyright are not
used to limit the access or legal rights of the compilation's users
beyond what the individual works permit. Inclusion of a covered work
in an aggregate does not cause this License to apply to the other
parts of the aggregate.
6. Conveying Non-Source Forms.
You may convey a covered work in object code form under the terms
of sections 4 and 5, provided that you also convey the
machine-readable Corresponding Source under the terms of this License,
in one of these ways:
a) Convey the object code in, or embodied in, a physical product
(including a physical distribution medium), accompanied by the
Corresponding Source fixed on a durable physical medium
customarily used for software interchange.
b) Convey the object code in, or embodied in, a physical product
(including a physical distribution medium), accompanied by a
written offer, valid for at least three years and valid for as
long as you offer spare parts or customer support for that product
model, to give anyone who possesses the object code either (1) a
copy of the Corresponding Source for all the software in the
product that is covered by this License, on a durable physical
medium customarily used for software interchange, for a price no
more than your reasonable cost of physically performing this
conveying of source, or (2) access to copy the
Corresponding Source from a network server at no charge.
c) Convey individual copies of the object code with a copy of the
written offer to provide the Corresponding Source. This
alternative is allowed only occasionally and noncommercially, and
only if you received the object code with such an offer, in accord
with subsection 6b.
d) Convey the object code by offering access from a designated
place (gratis or for a charge), and offer equivalent access to the
Corresponding Source in the same way through the same place at no
further charge. You need not require recipients to copy the
Corresponding Source along with the object code. If the place to
copy the object code is a network server, the Corresponding Source
may be on a different server (operated by you or a third party)
that supports equivalent copying facilities, provided you maintain
clear directions next to the object code saying where to find the
Corresponding Source. Regardless of what server hosts the
Corresponding Source, you remain obligated to ensure that it is
available for as long as needed to satisfy these requirements.
e) Convey the object code using peer-to-peer transmission, provided
you inform other peers where the object code and Corresponding
Source of the work are being offered to the general public at no
charge under subsection 6d.
A separable portion of the object code, whose source code is excluded
from the Corresponding Source as a System Library, need not be
included in conveying the object code work.
A "User Product" is either (1) a "consumer product", which means any
tangible personal property which is normally used for personal, family,
or household purposes, or (2) anything designed or sold for incorporation
into a dwelling. In determining whether a product is a consumer product,
doubtful cases shall be resolved in favor of coverage. For a particular
product received by a particular user, "normally used" refers to a
typical or common use of that class of product, regardless of the status
of the particular user or of the way in which the particular user
actually uses, or expects or is expected to use, the product. A product
is a consumer product regardless of whether the product has substantial
commercial, industrial or non-consumer uses, unless such uses represent
the only significant mode of use of the product.
"Installation Information" for a User Product means any methods,
procedures, authorization keys, or other information required to install
and execute modified versions of a covered work in that User Product from
a modified version of its Corresponding Source. The information must
suffice to ensure that the continued functioning of the modified object
code is in no case prevented or interfered with solely because
modification has been made.
If you convey an object code work under this section in, or with, or
specifically for use in, a User Product, and the conveying occurs as
part of a transaction in which the right of possession and use of the
User Product is transferred to the recipient in perpetuity or for a
fixed term (regardless of how the transaction is characterized), the
Corresponding Source conveyed under this section must be accompanied
by the Installation Information. But this requirement does not apply
if neither you nor any third party retains the ability to install
modified object code on the User Product (for example, the work has
been installed in ROM).
The requirement to provide Installation Information does not include a
requirement to continue to provide support service, warranty, or updates
for a work that has been modified or installed by the recipient, or for
the User Product in which it has been modified or installed. Access to a
network may be denied when the modification itself materially and
adversely affects the operation of the network or violates the rules and
protocols for communication across the network.
Corresponding Source conveyed, and Installation Information provided,
in accord with this section must be in a format that is publicly
documented (and with an implementation available to the public in
source code form), and must require no special password or key for
unpacking, reading or copying.
7. Additional Terms.
"Additional permissions" are terms that supplement the terms of this
License by making exceptions from one or more of its conditions.
Additional permissions that are applicable to the entire Program shall
be treated as though they were included in this License, to the extent
that they are valid under applicable law. If additional permissions
apply only to part of the Program, that part may be used separately
under those permissions, but the entire Program remains governed by
this License without regard to the additional permissions.
When you convey a copy of a covered work, you may at your option
remove any additional permissions from that copy, or from any part of
it. (Additional permissions may be written to require their own
removal in certain cases when you modify the work.) You may place
additional permissions on material, added by you to a covered work,
for which you have or can give appropriate copyright permission.
Notwithstanding any other provision of this License, for material you
add to a covered work, you may (if authorized by the copyright holders of
that material) supplement the terms of this License with terms:
a) Disclaiming warranty or limiting liability differently from the
terms of sections 15 and 16 of this License; or
b) Requiring preservation of specified reasonable legal notices or
author attributions in that material or in the Appropriate Legal
Notices displayed by works containing it; or
c) Prohibiting misrepresentation of the origin of that material, or
requiring that modified versions of such material be marked in
reasonable ways as different from the original version; or
d) Limiting the use for publicity purposes of names of licensors or
authors of the material; or
e) Declining to grant rights under trademark law for use of some
trade names, trademarks, or service marks; or
f) Requiring indemnification of licensors and authors of that
material by anyone who conveys the material (or modified versions of
it) with contractual assumptions of liability to the recipient, for
any liability that these contractual assumptions directly impose on
those licensors and authors.
All other non-permissive additional terms are considered "further
restrictions" within the meaning of section 10. If the Program as you
received it, or any part of it, contains a notice stating that it is
governed by this License along with a term that is a further
restriction, you may remove that term. If a license document contains
a further restriction but permits relicensing or conveying under this
License, you may add to a covered work material governed by the terms
of that license document, provided that the further restriction does
not survive such relicensing or conveying.
If you add terms to a covered work in accord with this section, you
must place, in the relevant source files, a statement of the
additional terms that apply to those files, or a notice indicating
where to find the applicable terms.
Additional terms, permissive or non-permissive, may be stated in the
form of a separately written license, or stated as exceptions;
the above requirements apply either way.
8. Termination.
You may not propagate or modify a covered work except as expressly
provided under this License. Any attempt otherwise to propagate or
modify it is void, and will automatically terminate your rights under
this License (including any patent licenses granted under the third
paragraph of section 11).
However, if you cease all violation of this License, then your
license from a particular copyright holder is reinstated (a)
provisionally, unless and until the copyright holder explicitly and
finally terminates your license, and (b) permanently, if the copyright
holder fails to notify you of the violation by some reasonable means
prior to 60 days after the cessation.
Moreover, your license from a particular copyright holder is
reinstated permanently if the copyright holder notifies you of the
violation by some reasonable means, this is the first time you have
received notice of violation of this License (for any work) from that
copyright holder, and you cure the violation prior to 30 days after
your receipt of the notice.
Termination of your rights under this section does not terminate the
licenses of parties who have received copies or rights from you under
this License. If your rights have been terminated and not permanently
reinstated, you do not qualify to receive new licenses for the same
material under section 10.
9. Acceptance Not Required for Having Copies.
You are not required to accept this License in order to receive or
run a copy of the Program. Ancillary propagation of a covered work
occurring solely as a consequence of using peer-to-peer transmission
to receive a copy likewise does not require acceptance. However,
nothing other than this License grants you permission to propagate or
modify any covered work. These actions infringe copyright if you do
not accept this License. Therefore, by modifying or propagating a
covered work, you indicate your acceptance of this License to do so.
10. Automatic Licensing of Downstream Recipients.
Each time you convey a covered work, the recipient automatically
receives a license from the original licensors, to run, modify and
propagate that work, subject to this License. You are not responsible
for enforcing compliance by third parties with this License.
An "entity transaction" is a transaction transferring control of an
organization, or substantially all assets of one, or subdividing an
organization, or merging organizations. If propagation of a covered
work results from an entity transaction, each party to that
transaction who receives a copy of the work also receives whatever
licenses to the work the party's predecessor in interest had or could
give under the previous paragraph, plus a right to possession of the
Corresponding Source of the work from the predecessor in interest, if
the predecessor has it or can get it with reasonable efforts.
You may not impose any further restrictions on the exercise of the
rights granted or affirmed under this License. For example, you may
not impose a license fee, royalty, or other charge for exercise of
rights granted under this License, and you may not initiate litigation
(including a cross-claim or counterclaim in a lawsuit) alleging that
any patent claim is infringed by making, using, selling, offering for
sale, or importing the Program or any portion of it.
11. Patents.
A "contributor" is a copyright holder who authorizes use under this
License of the Program or a work on which the Program is based. The
work thus licensed is called the contributor's "contributor version".
A contributor's "essential patent claims" are all patent claims
owned or controlled by the contributor, whether already acquired or
hereafter acquired, that would be infringed by some manner, permitted
by this License, of making, using, or selling its contributor version,
but do not include claims that would be infringed only as a
consequence of further modification of the contributor version. For
purposes of this definition, "control" includes the right to grant
patent sublicenses in a manner consistent with the requirements of
this License.
Each contributor grants you a non-exclusive, worldwide, royalty-free
patent license under the contributor's essential patent claims, to
make, use, sell, offer for sale, import and otherwise run, modify and
propagate the contents of its contributor version.
In the following three paragraphs, a "patent license" is any express
agreement or commitment, however denominated, not to enforce a patent
(such as an express permission to practice a patent or covenant not to
sue for patent infringement). To "grant" such a patent license to a
party means to make such an agreement or commitment not to enforce a
patent against the party.
If you convey a covered work, knowingly relying on a patent license,
and the Corresponding Source of the work is not available for anyone
to copy, free of charge and under the terms of this License, through a
publicly available network server or other readily accessible means,
then you must either (1) cause the Corresponding Source to be so
available, or (2) arrange to deprive yourself of the benefit of the
patent license for this particular work, or (3) arrange, in a manner
consistent with the requirements of this License, to extend the patent
license to downstream recipients. "Knowingly relying" means you have
actual knowledge that, but for the patent license, your conveying the
covered work in a country, or your recipient's use of the covered work
in a country, would infringe one or more identifiable patents in that
country that you have reason to believe are valid.
If, pursuant to or in connection with a single transaction or
arrangement, you convey, or propagate by procuring conveyance of, a
covered work, and grant a patent license to some of the parties
receiving the covered work authorizing them to use, propagate, modify
or convey a specific copy of the covered work, then the patent license
you grant is automatically extended to all recipients of the covered
work and works based on it.
A patent license is "discriminatory" if it does not include within
the scope of its coverage, prohibits the exercise of, or is
conditioned on the non-exercise of one or more of the rights that are
specifically granted under this License. You may not convey a covered
work if you are a party to an arrangement with a third party that is
in the business of distributing software, under which you make payment
to the third party based on the extent of your activity of conveying
the work, and under which the third party grants, to any of the
parties who would receive the covered work from you, a discriminatory
patent license (a) in connection with copies of the covered work
conveyed by you (or copies made from those copies), or (b) primarily
for and in connection with specific products or compilations that
contain the covered work, unless you entered into that arrangement,
or that patent license was granted, prior to 28 March 2007.
Nothing in this License shall be construed as excluding or limiting
any implied license or other defenses to infringement that may
otherwise be available to you under applicable patent law.
12. No Surrender of Others' Freedom.
If conditions are imposed on you (whether by court order, agreement or
otherwise) that contradict the conditions of this License, they do not
excuse you from the conditions of this License. If you cannot convey a
covered work so as to satisfy simultaneously your obligations under this
License and any other pertinent obligations, then as a consequence you may
not convey it at all. For example, if you agree to terms that obligate you
to collect a royalty for further conveying from those to whom you convey
the Program, the only way you could satisfy both those terms and this
License would be to refrain entirely from conveying the Program.
13. Remote Network Interaction; Use with the GNU General Public License.
Notwithstanding any other provision of this License, if you modify the
Program, your modified version must prominently offer all users
interacting with it remotely through a computer network (if your version
supports such interaction) an opportunity to receive the Corresponding
Source of your version by providing access to the Corresponding Source
from a network server at no charge, through some standard or customary
means of facilitating copying of software. This Corresponding Source
shall include the Corresponding Source for any work covered by version 3
of the GNU General Public License that is incorporated pursuant to the
following paragraph.
Notwithstanding any other provision of this License, you have
permission to link or combine any covered work with a work licensed
under version 3 of the GNU General Public License into a single
combined work, and to convey the resulting work. The terms of this
License will continue to apply to the part which is the covered work,
but the work with which it is combined will remain governed by version
3 of the GNU General Public License.
14. Revised Versions of this License.
The Free Software Foundation may publish revised and/or new versions of
the GNU Affero General Public License from time to time. Such new versions
will be similar in spirit to the present version, but may differ in detail to
address new problems or concerns.
Each version is given a distinguishing version number. If the
Program specifies that a certain numbered version of the GNU Affero General
Public License "or any later version" applies to it, you have the
option of following the terms and conditions either of that numbered
version or of any later version published by the Free Software
Foundation. If the Program does not specify a version number of the
GNU Affero General Public License, you may choose any version ever published
by the Free Software Foundation.
If the Program specifies that a proxy can decide which future
versions of the GNU Affero General Public License can be used, that proxy's
public statement of acceptance of a version permanently authorizes you
to choose that version for the Program.
Later license versions may give you additional or different
permissions. However, no additional obligations are imposed on any
author or copyright holder as a result of your choosing to follow a
later version.
15. Disclaimer of Warranty.
THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF
ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
16. Limitation of Liability.
IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
SUCH DAMAGES.
17. Interpretation of Sections 15 and 16.
If the disclaimer of warranty and limitation of liability provided
above cannot be given local legal effect according to their terms,
reviewing courts shall apply local law that most closely approximates
an absolute waiver of all civil liability in connection with the
Program, unless a warranty or assumption of liability accompanies a
copy of the Program in return for a fee.
END OF TERMS AND CONDITIONS
How to Apply These Terms to Your New Programs
If you develop a new program, and you want it to be of the greatest
possible use to the public, the best way to achieve this is to make it
free software which everyone can redistribute and change under these terms.
To do so, attach the following notices to the program. It is safest
to attach them to the start of each source file to most effectively
state the exclusion of warranty; and each file should have at least
the "copyright" line and a pointer to where the full notice is found.
<one line to give the program's name and a brief idea of what it does.>
Copyright (C) <year> <name of author>
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License
along with this program. If not, see <https://www.gnu.org/licenses/>.
Also add information on how to contact you by electronic and paper mail.
If your software can interact with users remotely through a computer
network, you should also make sure that it provides a way for users to
get its source. For example, if your program is a web application, its
interface could display a "Source" link that leads users to an archive
of the code. There are many ways you could offer source, and different
solutions will be better for different programs; see section 13 for the
specific requirements.
You should also get your employer (if you work as a programmer) or school,
if any, to sign a "copyright disclaimer" for the program, if necessary.
For more information on this, and how to apply and follow the GNU AGPL, see
<https://www.gnu.org/licenses/>.

4177
Cargo.lock generated Normal file
View file

File diff suppressed because it is too large Load diff

24
Cargo.toml Normal file
View file

@ -0,0 +1,24 @@
[package]
name = "little-librarian"
version = "0.1.0"
license = "AGPL-3.0"
edition = "2024"
[dependencies]
clap = { version = "4.5.40", features = ["derive"] }
futures = "0.3.31"
lopdf = "0.36.0"
quick-xml = "0.38.0"
rayon = "1.10.0"
ring = "0.17.14"
scraper = "0.23.1"
sha2 = "0.10.9"
snafu = { version = "0.8.6", features = ["rust_1_81"] }
sqlx = { version = "0.8.6", features = [ "runtime-tokio", "tls-rustls-ring", "migrate", "postgres", "derive" ] }
tokenizers = "0.21.2"
tokio = { version = "1.45.1", features = ["rt-multi-thread", "macros"]}
tracing = "0.1.41"
tracing-subscriber = "0.3.19"
tract-onnx = "0.21.13"
walkdir = "2.5.0"
zip = "4.2.0"

112
README.md Normal file
View file

@ -0,0 +1,112 @@
# Little Librarian
Semantic search for [Calibre](https://calibre-ebook.com/) libraries using vector
embeddings.
## Inner Workings
1. **Extract** text content from EPUB and PDF files
2. **Chunk** text into manageable pieces (500 words by default)
3. **Embed** each chunk using your ONNX model
4. **Store** embeddings in PostgreSQL with pgvector
5. **Search** by converting queries to embeddings and finding similar chunks
6. **Rerank** results using a cross-encoder model
## Shortcomings
This is an experiment and as such has some quite rough edges, including but not
limited to:
- Text chunking is naive, no idea if it even works assably on non-latin scripts
- PDF text extraction is hit-and-miss
- CPU only (this is also a strength, but at least the possibility of using GPU
would be nice)
## Requirements
All dependencies for running the application are specified in the
[flake.nix](./flake.nix) file within the project folder (look for the key
`buildInputs`, though [rustup](https://rustup.rs/) might just work with the
provided toolchain file). My recommendation is to use the provided
[nix flake](https://nixos.wiki/wiki/Flakes) directly, as it sets up everything
automatically.
Additional dependencies:
- [PostgreSQL](https://www.postgresql.org/) with
[pgvector extension](https://github.com/pgvector/pgvector)
- ONNX embedding model (e.g.,
[all-MiniLM-L6-v2](https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2))
- Tokenizer file (JSON format, from the embedding model)
- ONNX reranking model (e.g.
[bge-reranker-base](https://huggingface.co/BAAI/bge-reranker-base))
## Usage
Use `cargo run -- <ARGS>` to run directly from source code and specify needed
arguments in place of `<ARGS>`.
### Migrations
Either use [sqlx-cli](https://crates.io/crates/sqlx-cli) with `sqlx migrate run`
or apply the [migrations](./migrations/) manually.
### General
```
Usage: little-librarian [OPTIONS] <COMMAND>
Commands:
ask Configuration for the ask command
scan Configuration for the scan command
help Print this message or the help of the given subcommand(s)
Options:
-d, --db-url <DB_URL>
Where to read the database connection string from ('-' is stdin) [default: -]
-t, --tokenizer-path <TOKENIZER_PATH>
Path to content tokenizer [default: ./tokenizer.json]
-m, --model-path <MODEL_PATH>
Path to ONNX model used for embeddings.. [default: ./model.onnx]
-c, --chunk-size <CHUNK_SIZE>
Maximum text chunk size when splitting up content. Documents are not updated if chunk size changes [default: 500]
-h, --help
Print help
-V, --version
Print version
```
### Scan a Calibre library
This can take a huge amount of time, depending on the size of the library. It
will use up all available CPU cores.
```
Usage: little-librarian scan --library-path <LIBRARY_PATH>
Options:
-l, --library-path <LIBRARY_PATH> Root directory of calibre library
-h, --help Print help
-V, --version Print version
```
```
little-librarian scan --library-path /path/to/calibre/library
```
### Search for books
```
Usage: little-librarian ask [OPTIONS] --query <QUERY>
Options:
-q, --query <QUERY> Query to run
-l, --limit <LIMIT> Maximum number of results [default: 5]
-r, --reranking-path <RERANKING_PATH> Path to reranking model [default: ./reranking.onnx]
-h, --help Print help
-V, --version Print version
```
```
little-librarian ask --query "books about artificial intelligence ethics"
```

241
deny.toml Normal file
View file

@ -0,0 +1,241 @@
# This template contains all of the possible sections and their default values
# Note that all fields that take a lint level have these possible values:
# * deny - An error will be produced and the check will fail
# * warn - A warning will be produced, but the check will not fail
# * allow - No warning or error will be produced, though in some cases a note
# will be
# The values provided in this template are the default values that will be used
# when any section or field is not specified in your own configuration
# Root options
# The graph table configures how the dependency graph is constructed and thus
# which crates the checks are performed against
[graph]
# If 1 or more target triples (and optionally, target_features) are specified,
# only the specified targets will be checked when running `cargo deny check`.
# This means, if a particular package is only ever used as a target specific
# dependency, such as, for example, the `nix` crate only being used via the
# `target_family = "unix"` configuration, that only having windows targets in
# this list would mean the nix crate, as well as any of its exclusive
# dependencies not shared by any other crates, would be ignored, as the target
# list here is effectively saying which targets you are building for.
targets = [
# The triple can be any string, but only the target triples built in to
# rustc (as of 1.40) can be checked against actual config expressions
#"x86_64-unknown-linux-musl",
# You can also specify which target_features you promise are enabled for a
# particular target. target_features are currently not validated against
# the actual valid features supported by the target architecture.
#{ triple = "wasm32-unknown-unknown", features = ["atomics"] },
]
# When creating the dependency graph used as the source of truth when checks are
# executed, this field can be used to prune crates from the graph, removing them
# from the view of cargo-deny. This is an extremely heavy hammer, as if a crate
# is pruned from the graph, all of its dependencies will also be pruned unless
# they are connected to another crate in the graph that hasn't been pruned,
# so it should be used with care. The identifiers are [Package ID Specifications]
# (https://doc.rust-lang.org/cargo/reference/pkgid-spec.html)
#exclude = []
# If true, metadata will be collected with `--all-features`. Note that this can't
# be toggled off if true, if you want to conditionally enable `--all-features` it
# is recommended to pass `--all-features` on the cmd line instead
all-features = false
# If true, metadata will be collected with `--no-default-features`. The same
# caveat with `all-features` applies
no-default-features = false
# If set, these feature will be enabled when collecting metadata. If `--features`
# is specified on the cmd line they will take precedence over this option.
#features = []
# The output table provides options for how/if diagnostics are outputted
[output]
# When outputting inclusion graphs in diagnostics that include features, this
# option can be used to specify the depth at which feature edges will be added.
# This option is included since the graphs can be quite large and the addition
# of features from the crate(s) to all of the graph roots can be far too verbose.
# This option can be overridden via `--feature-depth` on the cmd line
feature-depth = 1
# This section is considered when running `cargo deny check advisories`
# More documentation for the advisories section can be found here:
# https://embarkstudios.github.io/cargo-deny/checks/advisories/cfg.html
[advisories]
# The path where the advisory databases are cloned/fetched into
#db-path = "$CARGO_HOME/advisory-dbs"
# The url(s) of the advisory databases to use
#db-urls = ["https://github.com/rustsec/advisory-db"]
# A list of advisory IDs to ignore. Note that ignored advisories will still
# output a note when they are encountered.
ignore = [
#"RUSTSEC-0000-0000",
#{ id = "RUSTSEC-0000-0000", reason = "you can specify a reason the advisory is ignored" },
#"a-crate-that-is-yanked@0.1.1", # you can also ignore yanked crate versions if you wish
#{ crate = "a-crate-that-is-yanked@0.1.1", reason = "you can specify why you are ignoring the yanked crate" },
]
# If this is true, then cargo deny will use the git executable to fetch advisory database.
# If this is false, then it uses a built-in git library.
# Setting this to true can be helpful if you have special authentication requirements that cargo-deny does not support.
# See Git Authentication for more information about setting up git authentication.
#git-fetch-with-cli = true
# This section is considered when running `cargo deny check licenses`
# More documentation for the licenses section can be found here:
# https://embarkstudios.github.io/cargo-deny/checks/licenses/cfg.html
[licenses]
# List of explicitly allowed licenses
# See https://spdx.org/licenses/ for list of possible licenses
# [possible values: any SPDX 3.11 short identifier (+ optional exception)].
allow = [
"AGPL-3.0",
"Apache-2.0",
"BSD-3-Clause",
"CDLA-Permissive-2.0",
"ISC",
"MIT",
"MPL-2.0",
"Unicode-3.0",
"Zlib"
]
# The confidence threshold for detecting a license from license text.
# The higher the value, the more closely the license text must be to the
# canonical license text of a valid SPDX license file.
# [possible values: any between 0.0 and 1.0].
confidence-threshold = 0.8
# Allow 1 or more licenses on a per-crate basis, so that particular licenses
# aren't accepted for every possible crate as with the normal allow list
exceptions = [
# Each entry is the crate and version constraint, and its specific allow
# list
#{ allow = ["Zlib"], crate = "adler32" },
]
# Some crates don't have (easily) machine readable licensing information,
# adding a clarification entry for it allows you to manually specify the
# licensing information
#[[licenses.clarify]]
# The package spec the clarification applies to
#crate = "ring"
# The SPDX expression for the license requirements of the crate
#expression = "MIT AND ISC AND OpenSSL"
# One or more files in the crate's source used as the "source of truth" for
# the license expression. If the contents match, the clarification will be used
# when running the license check, otherwise the clarification will be ignored
# and the crate will be checked normally, which may produce warnings or errors
# depending on the rest of your configuration
#license-files = [
# Each entry is a crate relative path, and the (opaque) hash of its contents
#{ path = "LICENSE", hash = 0xbd0eed23 }
#]
[licenses.private]
# If true, ignores workspace crates that aren't published, or are only
# published to private registries.
# To see how to mark a crate as unpublished (to the official registry),
# visit https://doc.rust-lang.org/cargo/reference/manifest.html#the-publish-field.
ignore = false
# One or more private registries that you might publish crates to, if a crate
# is only published to private registries, and ignore is true, the crate will
# not have its license(s) checked
registries = [
#"https://sekretz.com/registry
]
# This section is considered when running `cargo deny check bans`.
# More documentation about the 'bans' section can be found here:
# https://embarkstudios.github.io/cargo-deny/checks/bans/cfg.html
[bans]
# Lint level for when multiple versions of the same crate are detected
multiple-versions = "warn"
# Lint level for when a crate version requirement is `*`
wildcards = "allow"
# The graph highlighting used when creating dotgraphs for crates
# with multiple versions
# * lowest-version - The path to the lowest versioned duplicate is highlighted
# * simplest-path - The path to the version with the fewest edges is highlighted
# * all - Both lowest-version and simplest-path are used
highlight = "all"
# The default lint level for `default` features for crates that are members of
# the workspace that is being checked. This can be overridden by allowing/denying
# `default` on a crate-by-crate basis if desired.
workspace-default-features = "allow"
# The default lint level for `default` features for external crates that are not
# members of the workspace. This can be overridden by allowing/denying `default`
# on a crate-by-crate basis if desired.
external-default-features = "allow"
# List of crates that are allowed. Use with care!
allow = [
#"ansi_term@0.11.0",
#{ crate = "ansi_term@0.11.0", reason = "you can specify a reason it is allowed" },
]
# List of crates to deny
deny = [
#"ansi_term@0.11.0",
#{ crate = "ansi_term@0.11.0", reason = "you can specify a reason it is banned" },
# Wrapper crates can optionally be specified to allow the crate when it
# is a direct dependency of the otherwise banned crate
#{ crate = "ansi_term@0.11.0", wrappers = ["this-crate-directly-depends-on-ansi_term"] },
]
# List of features to allow/deny
# Each entry the name of a crate and a version range. If version is
# not specified, all versions will be matched.
#[[bans.features]]
#crate = "reqwest"
# Features to not allow
#deny = ["json"]
# Features to allow
#allow = [
# "rustls",
# "__rustls",
# "__tls",
# "hyper-rustls",
# "rustls",
# "rustls-pemfile",
# "rustls-tls-webpki-roots",
# "tokio-rustls",
# "webpki-roots",
#]
# If true, the allowed features must exactly match the enabled feature set. If
# this is set there is no point setting `deny`
#exact = true
# Certain crates/versions that will be skipped when doing duplicate detection.
skip = [
#"ansi_term@0.11.0",
#{ crate = "ansi_term@0.11.0", reason = "you can specify a reason why it can't be updated/removed" },
]
# Similarly to `skip` allows you to skip certain crates during duplicate
# detection. Unlike skip, it also includes the entire tree of transitive
# dependencies starting at the specified crate, up to a certain depth, which is
# by default infinite.
skip-tree = [
#"ansi_term@0.11.0", # will be skipped along with _all_ of its direct and transitive dependencies
#{ crate = "ansi_term@0.11.0", depth = 20 },
]
# This section is considered when running `cargo deny check sources`.
# More documentation about the 'sources' section can be found here:
# https://embarkstudios.github.io/cargo-deny/checks/sources/cfg.html
[sources]
# Lint level for what to happen when a crate from a crate registry that is not
# in the allow list is encountered
unknown-registry = "warn"
# Lint level for what to happen when a crate from a git repository that is not
# in the allow list is encountered
unknown-git = "warn"
# List of URLs for allowed crate registries. Defaults to the crates.io index
# if not specified. If it is specified but empty, no registries are allowed.
allow-registry = ["https://github.com/rust-lang/crates.io-index"]
# List of URLs for allowed Git repositories
allow-git = []
[sources.allow-org]
# github.com organizations to allow git sources for
github = []
# gitlab.com organizations to allow git sources for
gitlab = []
# bitbucket.org organizations to allow git sources for
bitbucket = []

114
flake.lock generated Normal file
View file

@ -0,0 +1,114 @@
{
"nodes": {
"fenix": {
"inputs": {
"nixpkgs": "nixpkgs",
"rust-analyzer-src": "rust-analyzer-src"
},
"locked": {
"lastModified": 1751265761,
"narHash": "sha256-VSbk7ppgFSqBxlsWtzIO52iT9TKkLJrmVYOIhQi13Kg=",
"owner": "nix-community",
"repo": "fenix",
"rev": "19c910fe3de1768e64e76b5ee6daa346e6b07410",
"type": "github"
},
"original": {
"owner": "nix-community",
"repo": "fenix",
"type": "github"
}
},
"flake-utils": {
"inputs": {
"systems": "systems"
},
"locked": {
"lastModified": 1731533236,
"narHash": "sha256-l0KFg5HjrsfsO/JpG+r7fRrqm12kzFHyUHqHCVpMMbI=",
"owner": "numtide",
"repo": "flake-utils",
"rev": "11707dc2f618dd54ca8739b309ec4fc024de578b",
"type": "github"
},
"original": {
"owner": "numtide",
"repo": "flake-utils",
"type": "github"
}
},
"nixpkgs": {
"locked": {
"lastModified": 1751011381,
"narHash": "sha256-krGXKxvkBhnrSC/kGBmg5MyupUUT5R6IBCLEzx9jhMM=",
"owner": "nixos",
"repo": "nixpkgs",
"rev": "30e2e2857ba47844aa71991daa6ed1fc678bcbb7",
"type": "github"
},
"original": {
"owner": "nixos",
"ref": "nixos-unstable",
"repo": "nixpkgs",
"type": "github"
}
},
"nixpkgs_2": {
"locked": {
"lastModified": 1751271578,
"narHash": "sha256-P/SQmKDu06x8yv7i0s8bvnnuJYkxVGBWLWHaU+tt4YY=",
"owner": "NixOS",
"repo": "nixpkgs",
"rev": "3016b4b15d13f3089db8a41ef937b13a9e33a8df",
"type": "github"
},
"original": {
"owner": "NixOS",
"ref": "nixos-unstable",
"repo": "nixpkgs",
"type": "github"
}
},
"root": {
"inputs": {
"fenix": "fenix",
"flake-utils": "flake-utils",
"nixpkgs": "nixpkgs_2"
}
},
"rust-analyzer-src": {
"flake": false,
"locked": {
"lastModified": 1751210071,
"narHash": "sha256-v7XmmLBNRMzRXiVCeH60ZeGEqo+aagmwawI0Z9+EoXY=",
"owner": "rust-lang",
"repo": "rust-analyzer",
"rev": "6df12139bccaaeecf6a34789e0ca799d1fe99c53",
"type": "github"
},
"original": {
"owner": "rust-lang",
"ref": "nightly",
"repo": "rust-analyzer",
"type": "github"
}
},
"systems": {
"locked": {
"lastModified": 1681028828,
"narHash": "sha256-Vy1rq5AaRuLzOxct8nz4T6wlgyUR7zLU309k9mBC768=",
"owner": "nix-systems",
"repo": "default",
"rev": "da67096a3b9bf56a91d16901293e51ba5b49a27e",
"type": "github"
},
"original": {
"owner": "nix-systems",
"repo": "default",
"type": "github"
}
}
},
"root": "root",
"version": 7
}

38
flake.nix Normal file
View file

@ -0,0 +1,38 @@
{
inputs = {
nixpkgs.url = "github:NixOS/nixpkgs/nixos-unstable";
flake-utils.url = "github:numtide/flake-utils";
fenix.url = "github:nix-community/fenix";
};
outputs =
{
nixpkgs,
fenix,
flake-utils,
...
}:
flake-utils.lib.eachDefaultSystem (
system:
let
pkgs = import nixpkgs { inherit system; };
fx = fenix.packages.${system};
rust = fx.combine [
(fx.fromToolchainFile {
file = ./rust-toolchain.toml;
sha256 = "sha256-AJ6LX/Q/Er9kS15bn9iflkUwcgYqRQxiOIL2ToVAXaU=";
})
];
buildInputs = [
rust
pkgs.cargo-deny
pkgs.sqlx-cli
];
in
{
devShells.default = pkgs.mkShell {
buildInputs = buildInputs;
};
}
);
}

11
migrations/20241107141344_documents.sql Normal file
View file

@ -0,0 +1,11 @@
CREATE TABLE documents (
id BIGINT PRIMARY KEY GENERATED ALWAYS AS IDENTITY,
book_id BIGINT NOT NULL,
version INTEGER NOT NULL,
text_chunk TEXT NOT NULL,
embedding vector(384) NOT NULL
);
CREATE INDEX ON documents
USING ivfflat (embedding vector_cosine_ops)
WITH (lists = 100);

4
migrations/20250522141158_hashes.sql Normal file
View file

@ -0,0 +1,4 @@
CREATE TABLE hashes (
book_id BIGINT PRIMARY KEY,
hash BYTEA
);

4
rust-toolchain.toml Normal file
View file

@ -0,0 +1,4 @@
[toolchain]
channel = "1.85.0"
components = ["rust-src", "rustc", "cargo", "rust-std", "rustfmt", "clippy", "rust-analyzer"]
profile = "minimal"

170
src/calibre.rs Normal file
View file

@ -0,0 +1,170 @@
//! Calibre library book discovery and metadata extraction.
use std::{
fs::{self, File},
io::{self, Read},
num::ParseIntError,
path::{Path, PathBuf},
};
use quick_xml::{Reader, encoding::EncodingError, events::Event};
use ring::digest::{Context, Digest, SHA256};
use snafu::{ResultExt, Snafu};
use walkdir::WalkDir;
/// A book found in a Calibre library.
#[derive(Debug)]
pub struct Book {
/// Calibre book ID from metadata.
pub id: i64,
/// File extension (epub, pdf, etc).
pub extension: String,
/// Full path to the book file.
pub path: PathBuf,
/// SHA256 hash of the book file.
pub hash: Vec<u8>,
}
/// Iterator over books in a Calibre library directory structure.
pub struct BookIterator {
/// Directory walker for traversing the library.
walker: walkdir::IntoIter,
}
/// Error when computing file hash.
#[derive(Debug, Snafu)]
#[snafu(display("Failed to delete old document versions."))]
pub struct HashError {
source: io::Error,
}
/// Errors that occur when reading Calibre ID from OPF metadata.
#[derive(Debug, Snafu)]
pub enum ReadCalibreIdError {
#[snafu(display("Failed to read opf xml."))]
ReadXml { source: quick_xml::Error },
#[snafu(display("Failed to read xml text event."))]
ReadXmlText { source: quick_xml::Error },
#[snafu(display("Failed to decode xml text."))]
DecodeXml { source: EncodingError },
#[snafu(display("Failed to parse calibre id as integer."))]
ParseId { source: ParseIntError },
#[snafu(display("Failed to find calibre id in opf."))]
CalibreIdMissing,
}
/// Error when iterating through library directories.
#[derive(Debug, Snafu)]
#[snafu(display("Failed to walk the library tree."))]
pub struct IterateError {
source: walkdir::Error,
}
impl BookIterator {
/// Create a new iterator for the given Calibre library root directory.
pub fn new(root_path: &PathBuf) -> Self {
Self {
walker: WalkDir::new(root_path)
.follow_links(true)
.max_depth(3) // root/author/book/files
.into_iter(),
}
}
/// Check if a file path represents a supported ebook format.
fn is_ebook(path: &Path) -> bool {
let extensions = ["epub", "pdf"];
path.extension()
.and_then(|ext| ext.to_str())
.map(|ext| extensions.contains(&ext.to_lowercase().as_str()))
.unwrap_or(false)
}
}
/// Compute SHA256 hash of data from a reader.
fn sha256_digest<R: Read>(mut reader: R) -> Result<Digest, HashError> {
let mut context = Context::new(&SHA256);
let mut buffer = [0; 1024];
loop {
let count = reader.read(&mut buffer).context(HashSnafu)?;
if count == 0 {
break;
}
context.update(&buffer[..count]);
}
Ok(context.finish())
}
/// Extract the Calibre book ID from OPF metadata XML.
fn read_calibre_id(opf: &str) -> Result<i64, ReadCalibreIdError> {
let mut reader = Reader::from_str(opf);
reader.config_mut().trim_text(true);
loop {
match reader.read_event() {
Err(e) => return Err(e).context(ReadXmlSnafu),
Ok(Event::Eof) => break,
Ok(Event::Start(e)) => {
if e.name().as_ref() == b"dc:identifier"
&& e.attributes().any(|attr| {
attr.ok()
.filter(|a| {
a.key.as_ref() == b"id" && a.value.as_ref() == b"calibre_id"
})
.is_some()
})
{
if let Event::Text(text) = reader.read_event().context(ReadXmlTextSnafu)? {
let calibre_id = text.decode().context(DecodeXmlSnafu)?.into_owned();
let calibre_id = calibre_id.parse::<i64>().context(ParseIdSnafu)?;
return Ok(calibre_id);
}
}
}
_ => (),
}
}
CalibreIdMissingSnafu.fail()
}
impl Iterator for BookIterator {
type Item = Result<Book, IterateError>;
fn next(&mut self) -> Option<Self::Item> {
loop {
match self.walker.next()? {
Ok(entry) => {
if entry.file_type().is_dir() {
let path = entry.path().to_owned();
let opf_path = path.join("metadata.opf");
if opf_path.exists() {
let files = fs::read_dir(&path).ok()?;
for f in files.flatten() {
let f = f.path();
if BookIterator::is_ebook(&f) {
let opf = fs::read_to_string(opf_path).ok()?;
let id = read_calibre_id(&opf).ok()?;
let book = File::open(&f).ok()?;
let hash = sha256_digest(book).ok()?.as_ref().to_vec();
return Some(Ok(Book {
id,
extension: f.extension()?.to_string_lossy().to_string(),
path: f,
hash,
}));
}
}
}
}
}
Err(e) => return Some(Err(e).context(IterateSnafu)),
}
}
}
}

62
src/cli.rs Normal file
View file

@ -0,0 +1,62 @@
//! Command-line interface definitions.
use std::path::PathBuf;
use clap::{Args, Parser, Subcommand};
/// Main CLI configuration.
#[derive(Parser)]
#[command(version, about, long_about = None)]
#[command(propagate_version = true)]
pub struct Cli {
/// Where to read the database connection string from ('-' is stdin).
#[arg(short, long, default_value = "-")]
pub db_url: String,
/// Path to content tokenizer.
#[arg(short, long, default_value = "./tokenizer.json")]
pub tokenizer_path: PathBuf,
/// Path to ONNX model used for embeddings..
#[arg(short, long, default_value = "./model.onnx")]
pub model_path: PathBuf,
/// Maximum text chunk size when splitting up content. Documents are not updated if chunk size
/// changes.
#[arg(short, long, default_value = "500")]
pub chunk_size: usize,
#[command(subcommand)]
pub command: Commands,
}
/// Configuration for the ask command.
#[derive(Args, Clone)]
pub struct AskConfig {
/// Query to run.
#[arg(short, long)]
pub query: String,
/// Maximum number of results.
#[arg(short, long, default_value = "5")]
pub limit: usize,
/// Path to reranking model.
#[arg(short, long, default_value = "./reranking.onnx")]
pub reranking_path: PathBuf,
}
/// Configuration for the scan command.
#[derive(Args, Clone)]
pub struct ScanConfig {
/// Root directory of calibre library.
#[arg(short, long)]
pub library_path: PathBuf,
}
/// Available CLI commands.
#[derive(Subcommand)]
pub enum Commands {
Ask(AskConfig),
Scan(ScanConfig),
}

48
src/extractors/epub.rs Normal file
View file

@ -0,0 +1,48 @@
use std::io::{self, Cursor, Read};
use scraper::Html;
use snafu::{ResultExt, Snafu};
use zip::result::ZipError;
use super::extractor::{EpubSnafu, ExtractionError, Extractor, extension};
pub struct Epub;
#[derive(Debug, Snafu)]
pub enum EpubExtractionError {
#[snafu(display("Failed to open epub as zip archive."))]
OpenZip { source: ZipError },
#[snafu(display("Failed to open file inside epub archive."))]
ZipFileOpen { source: ZipError },
#[snafu(display("Failed to read file inside epub archive."))]
ZipFileRead { source: io::Error },
}
impl Extractor for Epub {
fn extract(input: &[u8]) -> Result<String, ExtractionError> {
let mut texts = Vec::new();
let input = Cursor::new(input);
let mut archive = zip::ZipArchive::new(input)
.context(OpenZipSnafu)
.context(EpubSnafu)?;
for i in 0..archive.len() {
let mut file = archive
.by_index(i)
.context(ZipFileOpenSnafu)
.context(EpubSnafu)?;
let extension = extension(file.enclosed_name().unwrap_or_default());
if extension == "xhtml" || extension == "html" || extension == "htm" {
let mut entry = String::new();
file.read_to_string(&mut entry)
.context(ZipFileReadSnafu)
.context(EpubSnafu)?;
let document = Html::parse_document(&entry);
let text: String = document.root_element().text().collect::<Vec<_>>().join(" ");
texts.push(text);
}
}
Ok(texts.join(" "))
}
}

41
src/extractors/extractor.rs Normal file
View file

@ -0,0 +1,41 @@
use std::{io::Read, path::Path};
use snafu::{ResultExt, Snafu};
use super::{
epub::{Epub, EpubExtractionError},
pdf::{Pdf, PdfExtractionError},
};
#[derive(Debug, Snafu)]
pub enum ExtractionError {
#[snafu(display("Failed to read input for extraction."))]
Read { source: std::io::Error },
#[snafu(display("Failed to extract pdf."), visibility(pub))]
Pdf { source: PdfExtractionError },
#[snafu(display("Failed to extract epub."), visibility(pub))]
Epub { source: EpubExtractionError },
}
pub trait Extractor {
fn extract(input: &[u8]) -> Result<String, ExtractionError>;
}
pub fn extension<P: AsRef<Path>>(path: P) -> String {
let path = path.as_ref();
path.extension()
.unwrap_or_default()
.to_string_lossy()
.into_owned()
}
pub fn extract<R: Read>(mut input: R, extension: &str) -> Result<Option<String>, ExtractionError> {
let mut buffer = Vec::new();
input.read_to_end(&mut buffer).context(ReadSnafu)?;
Ok(match extension {
"epub" => Some(Epub::extract(&buffer)?),
"pdf" => Some(Pdf::extract(&buffer)?),
_ => None,
})
}

30
src/extractors/pdf.rs Normal file
View file

@ -0,0 +1,30 @@
use lopdf::Document;
use snafu::{ResultExt, Snafu};
use super::extractor::{Extractor, PdfSnafu};
pub struct Pdf;
fn get_pdf_text(doc: &Document) -> Result<String, lopdf::Error> {
let pages: Vec<u32> = doc.get_pages().keys().cloned().collect();
let text = doc.extract_text(&pages)?;
Ok(text)
}
#[derive(Debug, Snafu)]
pub enum PdfExtractionError {
#[snafu(display("Failed to open pdf."))]
OpenPdf { source: lopdf::Error },
#[snafu(display("Failed to extract content from pdf."))]
ReadPdf { source: lopdf::Error },
}
impl Extractor for Pdf {
fn extract(input: &[u8]) -> Result<String, super::extractor::ExtractionError> {
let doc = Document::load_from(input)
.context(OpenPdfSnafu)
.context(PdfSnafu)?;
get_pdf_text(&doc).context(ReadPdfSnafu).context(PdfSnafu)
}
}

32
src/hash.rs Normal file
View file

@ -0,0 +1,32 @@
//! File hashing utilities.
use std::{
fs::File,
io::{self, Read},
path::Path,
};
use sha2::{Digest, Sha256};
/// Length of SHA256 hash in bytes.
pub const SHA256_LENGTH: usize = 32;
/// Compute SHA256 hash of a file.
pub fn file_sha256<P: AsRef<Path>>(path: P) -> io::Result<[u8; SHA256_LENGTH]> {
let mut file = File::open(path)?;
let mut hasher = Sha256::new();
let mut buffer = [0; 1024];
loop {
let bytes_read = file.read(&mut buffer)?;
if bytes_read == 0 {
break;
}
hasher.update(&buffer[..bytes_read]);
}
let hash = hasher.finalize();
let hash_bytes: [u8; SHA256_LENGTH] = hash.into();
Ok(hash_bytes)
}

13
src/lib.rs Normal file
View file

@ -0,0 +1,13 @@
pub mod calibre;
pub mod cli;
pub mod extractors {
pub mod epub;
pub mod extractor;
pub mod pdf;
}
pub mod hash;
pub mod query;
pub mod scanner;
pub mod storage;
pub mod text_encoder;
pub mod tokenize;

80
src/main.rs Normal file
View file

@ -0,0 +1,80 @@
//! Little Librarian - Semantic search for Calibre libraries.
use std::io;
use clap::Parser;
use little_librarian::{
cli::{Cli, Commands},
query,
scanner::{self},
storage,
text_encoder::{self, TextEncoder},
tokenize::{self, Tokenizer},
};
use snafu::{ResultExt, Snafu};
/// Top-level application errors.
#[derive(Debug, Snafu)]
pub enum Error {
#[snafu(display("Failed to read db-url from stdin."))]
ReadStdin { source: io::Error },
#[snafu(display("Failed to read db-url from file '{file}'."))]
ReadFile { source: io::Error, file: String },
#[snafu(display("Failed to load tokenizer."))]
TokenizerLoading { source: tokenize::FromFileError },
#[snafu(display("Failed to load embedding model."))]
ModelLoading {
source: text_encoder::NewFromFileError,
},
#[snafu(display("Failed to connect to database."))]
DbConnection { source: storage::ConnectionError },
#[snafu(display("Failed to process query."))]
Ask { source: query::AskError },
}
#[tokio::main]
#[snafu::report]
async fn main() -> Result<(), Error> {
tracing_subscriber::fmt::init();
let cli = Cli::parse();
let db_url = if cli.db_url == "-" {
io::read_to_string(io::stdin()).context(ReadStdinSnafu)?
} else {
std::fs::read_to_string(&cli.db_url).context(ReadFileSnafu { file: cli.db_url })?
};
let tokenizer = Tokenizer::from_file(&cli.tokenizer_path).context(TokenizerLoadingSnafu)?;
let embedder = TextEncoder::from_file(&cli.model_path).context(ModelLoadingSnafu)?;
let db = storage::Postgres::connect(&db_url, 10)
.await
.context(DbConnectionSnafu)?;
match &cli.command {
Commands::Ask(config) => {
query::ask(
&config.query,
&db,
&tokenizer,
&embedder,
&config.reranking_path,
cli.chunk_size,
config.limit,
)
.await
.context(AskSnafu)?;
}
Commands::Scan(config) => {
scanner::scan(
&db,
&config.library_path,
&tokenizer,
&embedder,
cli.chunk_size,
)
.await;
}
}
Ok(())
}

65
src/query.rs Normal file
View file

@ -0,0 +1,65 @@
//! Query processing and document retrieval.
use std::path::PathBuf;
use snafu::{ResultExt, Snafu};
use crate::{
storage::{self, Postgres},
text_encoder::{self, TextEncoder},
tokenize::{self, Tokenizer},
};
/// Errors that occur during query processing.
#[derive(Debug, Snafu)]
pub enum AskError {
#[snafu(display("Failed to encode query."))]
Encode { source: tokenize::EncodeError },
#[snafu(display("Failed to embed query."))]
Embed { source: text_encoder::EmbedError },
#[snafu(display("Failed to retrieve similar documents."))]
Query { source: storage::QueryError },
#[snafu(display("Failed to load reranker model."))]
LoadReranker {
source: text_encoder::NewFromFileError,
},
#[snafu(display("Failed to rerank documents."))]
Rerank { source: text_encoder::RerankError },
}
/// Process a user query and return ranked document matches.
pub async fn ask(
query: &str,
db: &Postgres,
tokenizer: &Tokenizer,
text_encoder: &TextEncoder,
reranker_path: &PathBuf,
chunk_size: usize,
limit: usize,
) -> Result<(), AskError> {
let encodings = tokenizer.encode(query, chunk_size).context(EncodeSnafu)?;
let embeddings = text_encoder
.embed(encodings[0].clone())
.context(EmbedSnafu)?;
let documents = db
.query(embeddings, (limit * 10) as i32)
.await
.context(QuerySnafu)?;
let reranker = TextEncoder::from_file(reranker_path).context(LoadRerankerSnafu)?;
let reranked_docs = reranker
.rerank(query, documents, tokenizer, limit)
.context(RerankSnafu)?;
for (i, doc) in reranked_docs.iter().enumerate() {
println!(
"{}. Book ID: {}, Score: {:.3}",
i + 1,
doc.book_id,
doc.similarity
);
println!(" {}\n", doc.text_chunk);
}
Ok(())
}

110
src/scanner.rs Normal file
View file

@ -0,0 +1,110 @@
//! Book scanning and processing for embedding generation.
use std::{fs::File, io, path::PathBuf};
use snafu::{ResultExt, Snafu};
use tracing::{error, info};
use crate::{
calibre::{Book, BookIterator},
extractors::{self, extractor::ExtractionError},
hash,
storage::{self, Postgres},
text_encoder::{self, TextEncoder},
tokenize::{self, Tokenizer},
};
/// Scan a Calibre library and process all books for embedding generation.
pub async fn scan(
db: &Postgres,
library_path: &PathBuf,
tokenizer: &Tokenizer,
text_encoder: &TextEncoder,
chunk_size: usize,
) {
let books = BookIterator::new(library_path);
for book_result in books {
match book_result {
Ok(book) => {
if let Err(e) = process_book(&book, db, tokenizer, text_encoder, chunk_size).await {
error!("Failed to process book {}: {e}", book.id);
}
}
Err(e) => {
error!("error fetching calibre book: {e}");
}
}
}
}
/// Errors that occur during book processing.
#[derive(Debug, Snafu)]
pub enum ProcessBookError {
#[snafu(display("Failed to hash book."))]
BookHash { source: io::Error },
#[snafu(display("Failed to check in db if document needs update."))]
UpdateCheck { source: storage::CheckUpdateError },
#[snafu(display("Failed to open book."))]
Open { source: io::Error },
#[snafu(display("Failed to extract document content."))]
Extraction { source: ExtractionError },
#[snafu(display("Failed to tokenize document content."))]
Encode { source: tokenize::EncodeError },
#[snafu(display("Failed to create embeddings from encodings."))]
Embedding { source: text_encoder::EmbedError },
#[snafu(display("Failed to look up document versions."))]
DocumentVersion { source: storage::VersionError },
#[snafu(display("Failed to store embeddings in database."))]
SaveEmbeddings { source: storage::EmbeddingError },
#[snafu(display("Failed to update document hash in database."))]
UpdateHash { source: storage::UpdateHashError },
#[snafu(display("Failed to delete old document versions in database."))]
RemoveOldVersions {
source: storage::RemoveOldVersionsError,
},
}
/// Process a single book: extract text, generate embeddings, and store in database.
async fn process_book(
book: &Book,
db: &Postgres,
tokenizer: &Tokenizer,
embedder: &TextEncoder,
chunk_size: usize,
) -> Result<(), ProcessBookError> {
let book_hash = hash::file_sha256(&book.path).context(BookHashSnafu)?;
if db
.book_needs_update(book.id, book_hash)
.await
.context(UpdateCheckSnafu)?
{
info!("processing {:?}", book.path);
let input = File::open(book.path.clone()).context(OpenSnafu)?;
if let Some(text) =
extractors::extractor::extract(&input, &book.extension).context(ExtractionSnafu)?
{
let encodings = tokenizer.encode(&text, chunk_size).context(EncodeSnafu)?;
let embeddings = embedder.embed_multiple(encodings).context(EmbeddingSnafu)?;
let version = db
.embedding_version(book.id)
.await
.context(DocumentVersionSnafu)?
+ 1;
db.add_multiple_embeddings(book.id, version, embeddings)
.await
.context(SaveEmbeddingsSnafu)?;
}
db.update_hash(book.id, book_hash)
.await
.context(UpdateHashSnafu)?;
db.remove_old_versions(book.id)
.await
.context(RemoveOldVersionsSnafu)?;
}
Ok(())
}

224
src/storage.rs Normal file
View file

@ -0,0 +1,224 @@
//! Database storage and retrieval for document embeddings.
use futures::future::try_join_all;
use snafu::{ResultExt, Snafu};
use sqlx::{FromRow, Pool, postgres::PgPoolOptions};
use crate::{hash::SHA256_LENGTH, text_encoder::Embeddings};
/// A document chunk with similarity score from vector search.
#[derive(Debug, FromRow)]
pub struct DocumentMatch {
/// Calibre book ID.
pub book_id: i64,
/// Text content of the chunk.
pub text_chunk: String,
/// Cosine similarity score (0.0 to 1.0).
pub similarity: f64,
}
/// PostgreSQL database connection pool and operations.
pub struct Postgres {
/// Connection pool for database operations.
pool: Pool<sqlx::Postgres>,
}
/// Error when connecting to database.
#[derive(Debug, Snafu)]
#[snafu(display("Failed to connect to database."))]
pub struct ConnectionError {
source: sqlx::Error,
}
/// Error when retrieving document version.
#[derive(Debug, Snafu)]
#[snafu(display("Failed to get highest document version."))]
pub struct VersionError {
source: sqlx::Error,
}
/// Error when storing embeddings.
#[derive(Debug, Snafu)]
#[snafu(display("Failed to store embeddings."))]
pub struct EmbeddingError {
source: sqlx::Error,
}
/// Error when querying for similar documents.
#[derive(Debug, Snafu)]
#[snafu(display("Failed to query similarity data."))]
pub struct QueryError {
source: sqlx::Error,
}
/// Error when updating document hash.
#[derive(Debug, Snafu)]
#[snafu(display("Failed to update document hash."))]
pub struct UpdateHashError {
source: sqlx::Error,
}
/// Error when checking if document needs update.
#[derive(Debug, Snafu)]
#[snafu(display("Failed to check if document needs update."))]
pub struct CheckUpdateError {
source: sqlx::Error,
}
/// Error when removing old document versions.
#[derive(Debug, Snafu)]
#[snafu(display("Failed to delete old document versions."))]
pub struct RemoveOldVersionsError {
source: sqlx::Error,
}
impl Postgres {
/// Create a new database connection pool.
pub async fn connect(db_url: &str, max_connections: u32) -> Result<Self, ConnectionError> {
let pool = PgPoolOptions::new()
.max_connections(max_connections)
.connect(db_url)
.await
.context(ConnectionSnafu)?;
Ok(Self { pool })
}
/// Get the highest version number for a book's embeddings.
pub async fn embedding_version(&self, book_id: i64) -> Result<i32, VersionError> {
let version = sqlx::query_scalar(
"SELECT COALESCE(MAX(version), 0) FROM documents WHERE book_id = $1",
)
.bind(book_id)
.fetch_one(&self.pool)
.await
.context(VersionSnafu)?;
Ok(version)
}
/// Store multiple embeddings for a book in parallel.
pub async fn add_multiple_embeddings(
&self,
book_id: i64,
version: i32,
embeddings: Vec<Embeddings>,
) -> Result<(), EmbeddingError> {
let inserts: Vec<_> = embeddings
.into_iter()
.map(|e| self.add_embeddings(book_id, version, e))
.collect();
try_join_all(inserts).await?;
Ok(())
}
/// Store a single embedding for a book.
pub async fn add_embeddings(
&self,
book_id: i64,
version: i32,
embeddings: Embeddings,
) -> Result<(), EmbeddingError> {
let vector_string = embeddings_to_vec_string(embeddings.0);
sqlx::query(
"INSERT INTO documents (book_id, version, embedding, text_chunk) VALUES ($1, $2, $3::vector, $4)",
)
.bind(book_id)
.bind(version)
.bind(vector_string)
.bind(embeddings.2)
.execute(&self.pool)
.await
.context(EmbeddingSnafu)?;
Ok(())
}
/// Find documents similar to the query embedding using cosine similarity.
pub async fn query(
&self,
query: Embeddings,
limit: i32,
) -> Result<Vec<DocumentMatch>, QueryError> {
let vector_string = embeddings_to_vec_string(query.0);
let books = sqlx::query_as::<_, DocumentMatch>(
"SELECT book_id, text_chunk, 1 - (embedding <=> $1::vector) as similarity
FROM documents
ORDER BY embedding <=> $1::vector
LIMIT $2",
)
.bind(vector_string)
.bind(limit)
.fetch_all(&self.pool)
.await
.context(QuerySnafu)?;
Ok(books)
}
/// Update or insert the hash for a book.
pub async fn update_hash(
&self,
book_id: i64,
hash: [u8; SHA256_LENGTH],
) -> Result<(), UpdateHashError> {
sqlx::query(
"INSERT INTO hashes (book_id, hash) VALUES ($1, $2)
ON CONFLICT (book_id) DO UPDATE SET hash = $2",
)
.bind(book_id)
.bind(&hash[..])
.execute(&self.pool)
.await
.context(UpdateHashSnafu)?;
Ok(())
}
/// Check if a book needs to be reprocessed based on its hash.
pub async fn book_needs_update(
&self,
book_id: i64,
hash: [u8; SHA256_LENGTH],
) -> Result<bool, CheckUpdateError> {
let exists = sqlx::query_scalar::<_, bool>(
"SELECT NOT EXISTS (
SELECT 1 FROM hashes
WHERE book_id = $1 AND hash = $2
)",
)
.bind(book_id)
.bind(&hash[..])
.fetch_one(&self.pool)
.await
.context(CheckUpdateSnafu)?;
Ok(exists)
}
/// Remove all but the latest version of embeddings for a book.
pub async fn remove_old_versions(&self, book_id: i64) -> Result<(), RemoveOldVersionsError> {
sqlx::query(
"DELETE FROM documents WHERE version < (SELECT COALESCE(MAX(version), 0) FROM documents WHERE book_id = $1)",
)
.bind(book_id)
.execute(&self.pool)
.await
.context(RemoveOldVersionsSnafu)?;
Ok(())
}
}
/// Convert a vector of floats to PostgreSQL vector format string.
fn embeddings_to_vec_string(vector: Vec<f32>) -> String {
format!(
"[{}]",
vector
.iter()
.map(|x| x.to_string())
.collect::<Vec<String>>()
.join(",")
)
}

228
src/text_encoder.rs Normal file
View file

@ -0,0 +1,228 @@
//! Text embedding and reranking generation using ONNX models.
use std::{
fs::File,
io::{self, Read},
path::Path,
sync::Arc,
};
use rayon::prelude::*;
use snafu::{ResultExt, Snafu};
use tract_onnx::prelude::*;
use crate::{
storage::DocumentMatch,
text_encoder::tract_ndarray::ShapeError,
tokenize::{self, Encoding, Tokenizer},
};
/// Compiled ONNX model type alias.
type Model = SimplePlan<
TypedFact,
Box<dyn TypedOp>,
tract_onnx::prelude::Graph<TypedFact, Box<dyn TypedOp>>,
>;
/// Embedding vector with optional offset and original text.
pub type Embeddings = (Vec<f32>, Option<usize>, String);
/// ONNX-based text embedder for generating vector representations.
pub struct TextEncoder {
/// Compiled ONNX model for inference.
model: Model,
}
/// Errors when loading text encoderfrom file.
#[derive(Debug, Snafu)]
pub enum NewFromFileError {
#[snafu(display("Failed to open model file."))]
OpenModel { source: io::Error },
#[snafu(display("Failed to instantiate model file."))]
Instantiate { source: NewError },
}
/// Errors when creating text encoder from model data.
#[derive(Debug, Snafu)]
pub enum NewError {
#[snafu(display("Failed to load model file."))]
Model { source: TractError },
#[snafu(display("Failed to optimize model."))]
Optimization { source: TractError },
#[snafu(display("Failed to convert model into runnable."))]
Runnable { source: TractError },
}
/// Error when converting input to tensor format.
#[derive(Debug, Snafu)]
#[snafu(display("Failed to convert inmput to i64 arc tensor."))]
pub struct ToArcTensorError {
source: ShapeError,
}
/// Errors that occur during embedding generation.
#[derive(Debug, Snafu)]
pub enum EmbedError {
#[snafu(display("Failed to run model."))]
Run { source: TractError },
#[snafu(display("Failed to transform results."))]
Transform { source: TractError },
#[snafu(display("Failed to pool results."))]
Pooling,
#[snafu(display("Failed to convert input."))]
EmbedConversion { source: ToArcTensorError },
}
/// Errors that occur during reranking.
#[derive(Debug, Snafu)]
pub enum RerankError {
#[snafu(display("Failed to tokenize query-document pair."))]
Tokenize { source: tokenize::EncodeError },
#[snafu(display("Failed to score pair."))]
Score { source: ScoreError },
}
/// Errors that occur during embedding generation.
#[derive(Debug, Snafu)]
pub enum ScoreError {
#[snafu(display("Failed to convert input."))]
RerankConversion { source: ToArcTensorError },
#[snafu(display("Failed to run model."))]
RunRerank { source: TractError },
#[snafu(display("Failed to transform results."))]
TransformRerank { source: TractError },
}
impl TextEncoder {
/// Create a text encoder from an ONNX model file.
pub fn from_file<P: AsRef<Path>>(path: P) -> Result<Self, NewFromFileError> {
let input = File::open(path).context(OpenModelSnafu)?;
Self::new(&input).context(InstantiateSnafu)
}
/// Create a text encoder from ONNX model data.
pub fn new<R: Read>(mut input: R) -> Result<Self, NewError> {
let model = tract_onnx::onnx()
.model_for_read(&mut input)
.context(ModelSnafu)?
.into_optimized()
.context(OptimizationSnafu)?
.into_runnable()
.context(RunnableSnafu)?;
Ok(Self { model })
}
/// Convert token IDs to i64 tensor format for model input.
fn to_i64_arc_tensor(
seq_length: usize,
input: &[u32],
) -> Result<Arc<Tensor>, ToArcTensorError> {
let input = Self::to_i64(input);
let tensors = tract_ndarray::Array2::from_shape_vec((1, seq_length), input)
.context(ToArcTensorSnafu)?
.into_arc_tensor();
Ok(tensors)
}
/// Convert u32 token IDs to i64 for model compatibility.
fn to_i64(input: &[u32]) -> Vec<i64> {
input.iter().map(|&id| id as i64).collect::<Vec<i64>>()
}
/// Generate embeddings for multiple text chunks in parallel.
pub fn embed_multiple(&self, encodings: Vec<Encoding>) -> Result<Vec<Embeddings>, EmbedError> {
encodings.into_par_iter().map(|e| self.embed(e)).collect()
}
/// Generate embedding vector for a single text chunk.
pub fn embed(&self, encoding: Encoding) -> Result<Embeddings, EmbedError> {
let seq_length = encoding.0.get_ids().len();
let input_ids = Self::to_i64_arc_tensor(seq_length, encoding.0.get_ids())
.context(EmbedConversionSnafu)?;
let attention_mask = Self::to_i64_arc_tensor(seq_length, encoding.0.get_attention_mask())
.context(EmbedConversionSnafu)?;
let token_type_ids = tract_ndarray::Array2::<i64>::zeros((1, seq_length)).into_arc_tensor();
let outputs = self
.model
.run(tvec!(
TValue::Const(input_ids),
TValue::Const(attention_mask),
TValue::Const(token_type_ids),
))
.context(RunSnafu)?;
let embeddings = outputs[0]
.to_array_view::<f32>()
.context(TransformSnafu)?
.into_owned();
let pooled = embeddings
.mean_axis(tract_ndarray::Axis(1))
.ok_or(PoolingSnafu.build())?;
// let embeddings = embeddings.into_raw_vec_and_offset();
let embeddings = pooled.into_raw_vec_and_offset();
Ok((embeddings.0, embeddings.1, encoding.1))
}
/// Rerank documents using cross-encoder model for better relevance scoring.
pub fn rerank(
&self,
query: &str,
documents: Vec<DocumentMatch>,
tokenizer: &Tokenizer,
limit: usize,
) -> Result<Vec<DocumentMatch>, RerankError> {
let mut scored_docs = Vec::new();
for doc in documents {
let pair_text = format!("{} [SEP] {}", query, doc.text_chunk);
let encodings = tokenizer.encode(&pair_text, 512).context(TokenizeSnafu)?;
let encoding = &encodings[0].0;
let score = self.score_pair(encoding).context(ScoreSnafu)?;
let mut updated_doc = doc;
updated_doc.similarity = score;
scored_docs.push(updated_doc);
}
scored_docs.sort_by(|a, b| {
b.similarity
.partial_cmp(&a.similarity)
.unwrap_or(std::cmp::Ordering::Equal)
});
Ok(scored_docs.into_iter().take(limit).collect())
}
/// Score a query-document pair using the reranking model.
fn score_pair(&self, encoding: &tokenizers::Encoding) -> Result<f64, ScoreError> {
let seq_length = encoding.get_ids().len();
let input_ids = Self::to_i64_arc_tensor(seq_length, encoding.get_ids())
.context(RerankConversionSnafu)?;
let attention_mask = Self::to_i64_arc_tensor(seq_length, encoding.get_attention_mask())
.context(RerankConversionSnafu)?;
let outputs = self
.model
.run(tvec!(
TValue::Const(input_ids),
TValue::Const(attention_mask),
))
.context(RunRerankSnafu)?;
// Reranking model outputs a single relevance score
let scores = outputs[0]
.to_array_view::<f32>()
.context(TransformRerankSnafu)?;
let score = scores[[0, 0]] as f64;
Ok(score)
}
}

83
src/tokenize.rs Normal file
View file

@ -0,0 +1,83 @@
//! Text tokenization and chunking for embedding models.
use std::path::Path;
use snafu::{ResultExt, Snafu};
/// Errors that occur when loading a tokenizer from file.
#[derive(Debug, Snafu)]
pub enum FromFileError {
#[snafu(display("Failed to read tokenizer file."))]
File { source: std::io::Error },
#[snafu(display("Failed to load tokenizer."))]
New { source: FromBytesError },
}
/// Error when creating tokenizer from bytes.
#[derive(Debug, Snafu)]
#[snafu(display("Failed to read tokenizer from file."))]
pub struct FromBytesError {
source: tokenizers::Error,
}
/// Errors that occur during text encoding.
#[derive(Debug, Snafu)]
pub enum EncodeError {
#[snafu(display("Failed to encode."))]
Encode { source: tokenizers::Error },
}
/// Wrapper around a tokenizer that handles text chunking.
pub struct Tokenizer {
/// The underlying tokenizer implementation.
inner: tokenizers::Tokenizer,
}
/// A tokenized encoding paired with its original text.
pub type Encoding = (tokenizers::Encoding, String);
impl Tokenizer {
/// Create a tokenizer from a file path.
pub fn from_file<P: AsRef<Path>>(path: P) -> Result<Self, FromFileError> {
let contents = std::fs::read(path).context(FileSnafu)?;
Self::new(&contents).context(NewSnafu)
}
/// Create a tokenizer from configuration bytes.
pub fn new<P: AsRef<[u8]>>(config: P) -> Result<Self, FromBytesError> {
let inner = tokenizers::Tokenizer::from_bytes(config).context(FromBytesSnafu)?;
Ok(Self { inner })
}
/// Normalize whitespace and remove extra spaces from text.
fn clean_text(text: &str) -> String {
text.chars()
.map(|c| match c {
'\n' | '\r' | '\t' => ' ',
c if c.is_whitespace() => ' ',
c => c,
})
.collect::<String>()
.split_whitespace()
.collect::<Vec<&str>>()
.join(" ")
}
/// Split text into word-based chunks and encode each chunk.
pub fn encode(&self, text: &str, chunk_size: usize) -> Result<Vec<Encoding>, EncodeError> {
let text = Self::clean_text(text);
let words: Vec<&str> = text.split_whitespace().collect();
words
.chunks(chunk_size)
.map(|word_chunk| word_chunk.join(" "))
.map(|chunk| {
let tokenizer_impl = &self.inner;
tokenizer_impl
.encode(chunk.clone(), false)
.context(EncodeSnafu)
.map(|e| (e, chunk))
})
.collect::<Result<Vec<Encoding>, EncodeError>>()
}
}