Commit 4d0dac87 authored by Jan Reimes's avatar Jan Reimes
Browse files

feat(ai): add --accelerate option for embedding backend selection 

- Introduced `--accelerate` flag in `ai process` to choose between
  sentence-transformers backends: torch, onnx, and openvino.
- Updated `AiConfig` to support `embedding_backend` with validation.
- Enhanced embedding pipeline to accommodate backend selection.
- Updated documentation to reflect new CLI options and configurations.
parent 0bcfa080

docs/ai.md

+7 −0
Original line number Diff line number Diff line
@@ -59,6 +59,7 @@ TDC_AI_LLM_API_BASE= # Optional: custom endpoint 


# Embedding Model (HuggingFace sentence-transformers)

TDC_AI_EMBEDDING_MODEL=sentence-transformers/all-MiniLM-L6-v2   # Default: popular 384-dim model

TDC_AI_EMBEDDING_BACKEND=torch                                  # torch | onnx | openvino (default: torch)



# Storage

TDC_AI_STORE_PATH=                             # Defaults to <cache_dir>/.ai/lancedb
@@ -161,6 +162,12 @@ tdoc-crawler ai workspace process -w my-project --force 


### 5. Single TDoc Operations



Process a single TDoc through the pipeline (classification, extraction, embeddings, graph). Use `--accelerate` to choose the sentence-transformers backend.



```bash

tdoc-crawler ai process --tdoc-id SP-240001 --accelerate onnx

```



______________________________________________________________________



## CLI Commands

docs/history.md

+1 −0
Original line number Diff line number Diff line
@@ -4,6 +4,7 @@ This document provides a chronological log of all significant changes and improv 


## Recent Changes



- **2026-03-06**: [AI embeddings accelerate backend option](history/2026-03-06_SUMMARY_01_AI_EMBEDDINGS_ACCELERATE_BACKEND.md)

- **2026-02-09**: [Align CLI options across commands](history/2026-02-09_SUMMARY_01_ALIGN_CLI_OPTIONS_ACROSS_COMMANDS.md)

- **2026-02-07**: [Spec download auto-crawl and bug fixes](history/2026-02-07_SUMMARY_01_SPEC_DOWNLOAD_AUTO_CRAWL_AND_BUG_FIXES.md)

- **2026-02-03**: [Lazy credential resolution](history/2026-02-03_SUMMARY_03_LAZY_CREDENTIAL_RESOLUTION.md)

docs/history/2026-03-06_SUMMARY_01_AI_EMBEDDINGS_ACCELERATE_BACKEND.md

0 → 100644
+27 −0
Original line number Diff line number Diff line 
# Summary - AI embeddings accelerate backend option



Added an `--accelerate` option for `ai process` so callers can select the sentence-transformers backend used for embedding generation (torch, onnx, openvino). This flows through configuration and the embedding pipeline to ensure consistent backend selection.



## Changes



### CLI



- Added `--accelerate/-a` to `ai process` with `torch | onnx | openvino` choices.

- Passed the selected backend through `AiConfig` into the pipeline.



### AI Configuration



- Added `embedding_backend` to `AiConfig` with validation and `TDC_AI_EMBEDDING_BACKEND` support.



### Embedding Pipeline



- Threaded backend selection into `EmbeddingsManager` creation and pipeline embedding stage.

- Fixed the placeholder backend argument when constructing `SentenceTransformer`.



### Dependencies



- Updated `tdoc-ai` dependencies to ensure all requested backends can be installed.



## Verification



- Not run (docs and code changes only).