Log the complete prompt, schema, extra_body content, full API response,
token counts, and full parsed JSON output. Add exc_info=True tracebacks
on all failure paths.
vLLM servers support JSON schema enforcement via extra_body (guided_json
or structured_outputs), not OpenAI's response_format protocol. LangChain's
with_structured_output(method='json_schema') sends response_format which
vLLM ignores, causing NoneType not iterable parsing errors.
- vLLM path: direct OpenAI SDK call with extra_body={guided_json|structured_outputs}
- OpenRouter path: unchanged with_structured_output(method='json_schema')
- Try new 'structured_outputs' format first, fall back to legacy 'guided_json'
- Update _SEED_DECOMPOSE with explicit JSON array instruction
- Add diagnostic logging: exc_info=True, schema preview, prompt template preview
- Add logging in _parse_legacy_json for fallback failure debugging
vLLM's chat_template_kwargs leaked into LangChain's AsyncCompletions.parse()
via _get_langchain_model's model_kwargs, causing structured decomposition
to fail on vLLM backends. Skip vLLM-specific params when building the
LangChain model — only provider-agnostic params (OpenAI reasoning) pass through.
- ChunkHighlightService.compute_highlights_batch(): single LLM call across
all cited chunks, grouped by sub-question, with structured output
- render_highlight_html(): self-contained HTML page with yellow-highlighted
relevant sentences, LLM reason annotations, and View Original PDF footer
- Per-target error isolation, ChromaDB miss handling, graceful degradation
- 14 tests: 7 batch service + 7 HTML rendering
The per-sub-question filter was all-or-nothing: if the LLM returned
9 scores for 10 chunks (common with qwen3.5-35b), every chunk was
discarded and the user got 'no relevant information found'.
Now: fewer scores → pad with 0.0; more scores → truncate. Changed
from error→warning since this is recoverable.
Also improve LTT page UI: sources collapsed by default in per-sub-q
sections, and the 'Your question' text now shows the full question
instead of being truncated.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Break the hardcoded per-sub-q filter prompt into 3 editable PromptService templates (filter_intro, filter_section, filter_outro) with placeholders for the for-loop iteration pattern. Refactor RelevanceFilter._build_per_subq_prompt() to compose them at runtime, falling back to built-in defaults when PromptService is unavailable.
Fix two latent bugs from Package 4:
- generate_per_subq was called by rag.py but never added to _VALID_STEPS or DB seed (would ValueError at runtime)
- _SEED_GENERATE placeholder mismatch: flat generate_response() expects {question}/{context} but Package 4 changed it to {context_sections}. Restored flat template; generate_per_subq now holds {context_sections}.
Add database backfill migration in seed_default_profiles() to INSERT OR IGNORE missing steps into existing profile rows, ensuring all 7 steps exist on restart.
Restructure System Prompts UI: remove unused flat filter/generate steps, replace with Step 2.1-2.3 (filter_intro/section/outro) and Step 3 (generate_per_subq). Update PlaceholderDocs with {context_sections}, {subq_idx}, {subq_question}.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Add chunks_retrieved_per_subq_count and chunks_filtered_per_subq_count columns to query_history table with safe ALTER TABLE migration. Replace generate template {question}/{context} placeholders with {context_sections} for per-sub-question organized context sections. Update Phase 3 test assertions to match new template and schema shapes.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Add retrieve_per_subquestion() that queries ChromaDB independently per sub-question instead of joining all sub-qs into one query string. Add filter_per_subquestion() that evaluates each chunk against its own originating sub-question in a single LLM call with a redesigned grouped prompt. Add generate_response_per_subquestion() that produces markdown sections per sub-question with grouped sources and {context_sections} template support. All existing methods preserved for backward compatibility.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Replace numeric [1] labels with [filename, page N] format in context chunks.
Update LLM prompt to instruct inline citation using bracket labels.
Enables traceable source references in generated answers.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Change QueryDecomposer prompt to generate 2-5 sub-questions instead of keywords. Rename API field from keywords to extracted_questions across models, service, and router.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
PDF uploads now use parse_pdf_by_page() -> chunk_pages() -> extract page PDFs -> enhanced metadata with page_number, chunk_file_path, and document_id. Same-filename replacement deletes old chunks and PDFs before re-ingest. DOCX/TXT keep original flat flow with document_id added. RAGService.ingest_document() accepts optional document_id parameter.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
- Add react-router-dom with NavBar component (LTT + RAG Database tabs)
- Extract AppContent into LTTPage, add RAGDatabasePage placeholder
- Refactor App.tsx to BrowserRouter + Routes layout
- Switch ResponsePanel to react-markdown for rich formatting
- Fix ResponsePanel test for markdown rendering
- Update RAG prompt to cite source name instead of number
- Save Phase 1 enhancement plan (.plans/phase1_enhancement_plan.md)
- Log extra_body contents before sending to LLM
- Log full LLM response object for debugging
- Changed extra_body format to OpenRouter reasoning format
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus \u003cclio-agent@sisyphuslabs.ai\u003e
The LLM (Qwen3.5 via OpenRouter) returns JSON wrapped in markdown code blocks:
```json
["project manager", "limits", ...]
```
But the code was trying to parse this directly with json.loads(), causing:
- QueryDecomposer to return empty keywords
- RelevanceFilter to fail with "Expecting value: line 1 column 1"
Changes:
- Added _extract_json_from_markdown() helper function to both modules
- Strips markdown code block markers (```json and ```) before JSON parsing
- Added unit tests for markdown code block handling
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus \u003cclio-agent@sisyphuslabs.ai\u003e
- LLMClient.complete() now accepts step_name parameter to identify processing step
- Logs prompt preview (first 100 + last 100 chars) at INFO level
- Logs processing time in milliseconds with token usage stats
- Updated QueryDecomposer, RelevanceFilter, and RAGService to pass step names
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Woody