lplalbert opened a new pull request, #417:
URL: https://github.com/apache/hugegraph-ai/pull/417

   Refs #107 (Tiny Task: fix TODO/FIXME in LLM code)
   
   ## What
   `rag_requests.py` baked `prompt.answer_prompt` / 
`prompt.keywords_extract_prompt` /
   `prompt.gremlin_generate_prompt` into the request models' `Query` defaults 
at import
   time (the file's own `# TODO: import the default value of prompt.* 
dynamically`
   admitted it), so the `req.x or prompt.x` fallbacks in `rag_api.py` never 
resolved the
   current `PromptConfig` for omitted fields — callers silently received the 
server-start
   snapshot even after a runtime prompt update. `KeywordExtract` carried the 
same bug
   class via its module-level `KEYWORDS_EXTRACT_TPL` snapshot.
   
   ## ⚠️ Behavior change (please weigh in)
   Omitted API prompt fields now resolve the **current** `PromptConfig` at 
request time
   instead of the import-time snapshot. Identical in practice unless prompts 
are updated
   at runtime (demo UI / `update_yaml_file()`) — such updates are now visible 
to API calls
   that omit the fields. Side effect: the huge prompt strings disappear from 
the OpenAPI
   field defaults (they now show `null`).
   
   ## Changes
   - `RAGRequest.answer_prompt/keywords_extract_prompt/gremlin_prompt`,
     `GraphRAGRequest.gremlin_prompt`, `GremlinGenerateRequest.gremlin_prompt` 
default to
     `None`; every call site resolves the live `prompt` object (`rag_api.py` 
fallbacks;
     `Text2GremlinNode`/`GremlinGenerateSynthesize` already fell back when 
given `None`).
   - `/text2gremlin` gets the same handler-level `req.x or prompt.x` fallback 
as `/rag`
     and `/rag/graph`, so all three endpoints resolve uniformly at request time.
   - `KeywordExtract` reads `prompt.keywords_extract_prompt` at construction 
instead of
     the removed import-time `KEYWORDS_EXTRACT_TPL` constant.
   - Tests in `test_prompt_config.py` (removes both FIXMEs): omitted fields 
resolve the
     current `PromptConfig` across `/rag`, `/rag/graph`, `/text2gremlin` and
     `KeywordExtract`; runtime-loaded `config_prompt.yaml` extract prompts pass 
the same
     example/parser contract as the class constants (EN + CN).
   
   ## Validation (actually run; offline only)
   - red→green (stash-demonstrated): the 4 resolution tests fail on the old 
defaults —
     failure output shows the import-time prompt strings directly — then 16 
passed after
     (whole `test_prompt_config.py`)
   - `SKIP_EXTERNAL_SERVICES=true uv run pytest src/tests/config/ src/tests/api/
     src/tests/operators/llm_op/ src/tests/nodes/` → 351 passed, 3 subtests 
passed,
     2 failed — the 2 failures are pre-existing on Windows 
(`test_external_config.py`
     asserts POSIX 0600 file-mode bits) and unrelated to this diff
   - `uv run --extra dev ruff format --check .` / `ruff check .` → clean (588 
files)
   - `uv run --extra dev ty check hugegraph-llm/src/hugegraph_llm` → 271 (same 
as
     current `main`)
   - Real-provider verification: none (offline tests only)
   
   Refs #107
   
   🤖 Generated with [Claude Code](https://claude.com/claude-code)
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to