awembed 0.1.0__tar.gz

awembed-0.1.0/LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

awembed-0.1.0/PKG-INFO

@@ -0,0 +1,126 @@
+Metadata-Version: 2.4
+Name: awembed
+Version: 0.1.0
+Summary: Train a small embedding model that knows your corpus: teacher capture, distillation, fidelity-gated int8, held-out eval -- one CLI, every stage gated.
+License: Apache-2.0
+Project-URL: Homepage, https://github.com/Aitherium/awembed
+Project-URL: Documentation, https://github.com/Aitherium/awembed#readme
+Project-URL: Repository, https://github.com/Aitherium/awembed.git
+Project-URL: Issues, https://github.com/Aitherium/awembed/issues
+Keywords: embeddings,distillation,retrieval,code-search,quantization
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: numpy>=1.24
+Requires-Dist: torch>=2.1
+Requires-Dist: transformers>=4.51
+Provides-Extra: teacher
+Requires-Dist: fastapi>=0.100; extra == "teacher"
+Requires-Dist: uvicorn>=0.23; extra == "teacher"
+Requires-Dist: einops; extra == "teacher"
+Requires-Dist: datasets; extra == "teacher"
+Requires-Dist: accelerate; extra == "teacher"
+Provides-Extra: dev
+Requires-Dist: pytest>=7.0; extra == "dev"
+Dynamic: license-file
+
+# awembed — Aither World Embed
+
+Train a small embedding model that knows **your** corpus, and prove it beats the big one.
+
+A 0.6B student trained this way on one codebase retrieved the right directory on the
+first try 80% of the time; the 7.85B general-purpose teacher it learned from managed 66%
+on the same held-out questions. The student ships as 1.06 GB of int8 with 0.999 cosine
+fidelity to its full-precision self. The whole run was one GPU for about two and a half
+hours.
+
+The reason a small model can win is not magic: the teacher never saw your corpus. The
+student is trained on two signals at once — the teacher's *margins* between the right
+answer and plausible wrong ones (distillation), and your corpus's own labels
+(in-domain contrastive learning). For retrieval over something you own, in-domain
+supervision is worth more than parameters.
+
+## Install
+
+```bash
+pip install awembed              # the student side: torch + transformers
+pip install "awembed[teacher]"   # also the teacher server (fastapi, einops, ...)
+```
+
+## The recipe
+
+```bash
+# 0. Build the corpus from per-directory READINGS of your repo: one JSON per line,
+#    {"rel": "<dir>", "reading": "<a paragraph on what the dir does>", "prompt_version": "v1"}.
+#    Any LLM over `ls` + a few file heads produces them (~1,500 chars each is plenty);
+#    a walker that writes them for you is the next thing this module grows.
+#    Output: one row per question, with the right directory and K hard negatives --
+#    directories that look plausible and are wrong -- split by DIRECTORY, so
+#    evaluation never sees a directory that training saw. Add your own
+#    question -> target pairs with --ground-truth (JSONL {"query", "target"}).
+awembed corpus --readings dir_readings.jsonl --out corpus.jsonl
+
+# 1. Before renting anything: does the teacher load and answer HERE?
+awembed probe                                  # exit 0 ok / 1 failed / 2 could not judge
+
+# 2. Capture the teacher's judgement for every row (starts the teacher server itself)
+awembed capture  --corpus corpus.jsonl --seed 1 --out artifacts/
+
+# 3. Distill the student (fp32 master + bf16 autocast; gate: the loss must fall)
+awembed distill  --corpus corpus.jsonl --seed 1 --out artifacts/
+
+# 4. Weight-only int8 export (gate: >= 0.98 mean cosine to the fp32 student)
+awembed quantize --corpus corpus.jsonl --seed 1 --out artifacts/
+
+# 5. Teacher vs untrained baseline vs student vs int8, on the held-out directories
+awembed eval     --corpus corpus.jsonl --seed 1 --out artifacts/
+
+# ...or all four artifact stages in order, stopping at the first gate that refuses:
+awembed run      --corpus corpus.jsonl --seed 1 --out artifacts/
+```
+
+Every stage writes a sidecar (`teacher_manifest.json`, `train_sidecar.json`,
+`quant_sidecar.json`, `eval_report.json`) with what it measured, so the record of a run
+is the run. Every stage has a `--self-test` that runs the same code path on a thumbnail
+model, on CPU, offline.
+
+## What the gates refuse
+
+| stage | refuses when |
+|---|---|
+| capture | a corpus row lacks inline hard-negative text (path-only negatives teach a shortcut, not retrieval) |
+| distill | the mean loss over the last 10 steps is not below the first 10, or any step is non-finite |
+| quantize | the int8 export embeds at < 0.98 mean cosine to the fp32 student, measured through the loader the eval stage uses |
+| eval | the student does not beat the baseline on both p@1 and recall@10, or int8 holds < 98% of the student's recall@10 |
+
+A run that "completed" with no learning is exactly the thing a training pipeline is
+best at hiding. The gates are the point.
+
+## Defaults, and what to change
+
+- **Teacher**: `nvidia/NV-Embed-v2` (7.85B, CC-BY-NC-4.0 — a teacher signal only; its
+  weights and its captured targets never ship). Any OpenAI-compatible `/v1/embeddings`
+  endpoint works: pass `--teacher-url` to `capture` instead of letting it start one.
+- **Student**: `Qwen/Qwen3-Embedding-0.6B` (Apache-2.0, 1024-dim, last-token pooling).
+  Pass `--student` to `distill` for another base.
+- **Asymmetry**: queries carry an instruction prefix; documents embed plain. The prefix
+  the student was trained with is the one you must use at query time. It is written in
+  the sidecar.
+- **Two environments**: the default teacher pins an older `transformers` than the
+  student needs, so run `teacher`/`probe` in their own venv and point `capture` at it
+  with `NV_EMBED_PYTHON=/path/to/venv/bin/python`.
+
+## Using the student
+
+The student is an ordinary Hugging Face directory (`artifacts/student/`) — serve it
+with vLLM (`--task embed`), Text Embeddings Inference, or `transformers` directly. It is
+a **new vector space** (1024-dim): index into a fresh collection, do not mix it with
+vectors from the model you replaced.
+
+It pairs with the rest of the Aither World family: `awgraph` (code graph search over
+your repo), `awm` (agent memory), `awfind` (ranked answers), `awrecurse` (documents
+larger than a context window), and `awdk` agents that consume any of those.
+
+## Licence
+
+Apache-2.0. Models you train carry their base model's licence.

awembed-0.1.0/README.md

@@ -0,0 +1,100 @@
+# awembed — Aither World Embed
+
+Train a small embedding model that knows **your** corpus, and prove it beats the big one.
+
+A 0.6B student trained this way on one codebase retrieved the right directory on the
+first try 80% of the time; the 7.85B general-purpose teacher it learned from managed 66%
+on the same held-out questions. The student ships as 1.06 GB of int8 with 0.999 cosine
+fidelity to its full-precision self. The whole run was one GPU for about two and a half
+hours.
+
+The reason a small model can win is not magic: the teacher never saw your corpus. The
+student is trained on two signals at once — the teacher's *margins* between the right
+answer and plausible wrong ones (distillation), and your corpus's own labels
+(in-domain contrastive learning). For retrieval over something you own, in-domain
+supervision is worth more than parameters.
+
+## Install
+
+```bash
+pip install awembed              # the student side: torch + transformers
+pip install "awembed[teacher]"   # also the teacher server (fastapi, einops, ...)
+```
+
+## The recipe
+
+```bash
+# 0. Build the corpus from per-directory READINGS of your repo: one JSON per line,
+#    {"rel": "<dir>", "reading": "<a paragraph on what the dir does>", "prompt_version": "v1"}.
+#    Any LLM over `ls` + a few file heads produces them (~1,500 chars each is plenty);
+#    a walker that writes them for you is the next thing this module grows.
+#    Output: one row per question, with the right directory and K hard negatives --
+#    directories that look plausible and are wrong -- split by DIRECTORY, so
+#    evaluation never sees a directory that training saw. Add your own
+#    question -> target pairs with --ground-truth (JSONL {"query", "target"}).
+awembed corpus --readings dir_readings.jsonl --out corpus.jsonl
+
+# 1. Before renting anything: does the teacher load and answer HERE?
+awembed probe                                  # exit 0 ok / 1 failed / 2 could not judge
+
+# 2. Capture the teacher's judgement for every row (starts the teacher server itself)
+awembed capture  --corpus corpus.jsonl --seed 1 --out artifacts/
+
+# 3. Distill the student (fp32 master + bf16 autocast; gate: the loss must fall)
+awembed distill  --corpus corpus.jsonl --seed 1 --out artifacts/
+
+# 4. Weight-only int8 export (gate: >= 0.98 mean cosine to the fp32 student)
+awembed quantize --corpus corpus.jsonl --seed 1 --out artifacts/
+
+# 5. Teacher vs untrained baseline vs student vs int8, on the held-out directories
+awembed eval     --corpus corpus.jsonl --seed 1 --out artifacts/
+
+# ...or all four artifact stages in order, stopping at the first gate that refuses:
+awembed run      --corpus corpus.jsonl --seed 1 --out artifacts/
+```
+
+Every stage writes a sidecar (`teacher_manifest.json`, `train_sidecar.json`,
+`quant_sidecar.json`, `eval_report.json`) with what it measured, so the record of a run
+is the run. Every stage has a `--self-test` that runs the same code path on a thumbnail
+model, on CPU, offline.
+
+## What the gates refuse
+
+| stage | refuses when |
+|---|---|
+| capture | a corpus row lacks inline hard-negative text (path-only negatives teach a shortcut, not retrieval) |
+| distill | the mean loss over the last 10 steps is not below the first 10, or any step is non-finite |
+| quantize | the int8 export embeds at < 0.98 mean cosine to the fp32 student, measured through the loader the eval stage uses |
+| eval | the student does not beat the baseline on both p@1 and recall@10, or int8 holds < 98% of the student's recall@10 |
+
+A run that "completed" with no learning is exactly the thing a training pipeline is
+best at hiding. The gates are the point.
+
+## Defaults, and what to change
+
+- **Teacher**: `nvidia/NV-Embed-v2` (7.85B, CC-BY-NC-4.0 — a teacher signal only; its
+  weights and its captured targets never ship). Any OpenAI-compatible `/v1/embeddings`
+  endpoint works: pass `--teacher-url` to `capture` instead of letting it start one.
+- **Student**: `Qwen/Qwen3-Embedding-0.6B` (Apache-2.0, 1024-dim, last-token pooling).
+  Pass `--student` to `distill` for another base.
+- **Asymmetry**: queries carry an instruction prefix; documents embed plain. The prefix
+  the student was trained with is the one you must use at query time. It is written in
+  the sidecar.
+- **Two environments**: the default teacher pins an older `transformers` than the
+  student needs, so run `teacher`/`probe` in their own venv and point `capture` at it
+  with `NV_EMBED_PYTHON=/path/to/venv/bin/python`.
+
+## Using the student
+
+The student is an ordinary Hugging Face directory (`artifacts/student/`) — serve it
+with vLLM (`--task embed`), Text Embeddings Inference, or `transformers` directly. It is
+a **new vector space** (1024-dim): index into a fresh collection, do not mix it with
+vectors from the model you replaced.
+
+It pairs with the rest of the Aither World family: `awgraph` (code graph search over
+your repo), `awm` (agent memory), `awfind` (ranked answers), `awrecurse` (documents
+larger than a context window), and `awdk` agents that consume any of those.
+
+## Licence
+
+Apache-2.0. Models you train carry their base model's licence.

awembed-0.1.0/awembed/__init__.py

@@ -0,0 +1,12 @@
+"""awembed -- Aither World Embed.
+
+Train a small embedding model that knows your corpus: capture a large teacher's
+judgement, distill it into a student together with your own labels, quantize with
+a fidelity gate, and evaluate on a split that holds out whole directories. Every
+stage is a standalone module with a `--self-test` and a gate that refuses an
+artifact that is not fit to hand on.
+"""
+
+__version__ = "0.1.0"
+
+__all__ = ["__version__"]