- Fix syntax error in server.py (dangling docstring lines)
- Correct model filename: flux-2-klein-4b.safetensors (without -fp8)
- Fix _WORKFLOW_REGISTRY key to match actual downloaded filename
- Update get_models() to always include registry models as fallback
- Fix test expectations to match corrected model names
- All 37 tests passing
conversation_chunks_fts is a standalone FTS5 table (no content= option).
The old INSERT ... VALUES('rebuild') is a no-op on standalone tables and
left deleted chunks searchable in the FTS shadow tables.
Fix: collect IDs before deletion, explicitly DELETE FROM conversation_chunks_fts
WHERE rowid IN (...) before removing from the main table. This keeps FTS
in sync after every vacuum call.
Tests: 303/303 passing. Vacuum tests now pass for the right reason.
- Add name (str) param: filename prefix saved as {name}_{timestamp}_{seed}.png
- Add count (int, 1-10) param: generate N images in one call
- Extract _sanitize_name() helper: strips special chars, collapses underscores, caps at 64 chars
- Extract _build_filename() helper: pure function for testable filename construction
- Extract _generate_single() coroutine: clean loop body for batch generation
- Fixed seed batches increment seed per image (seed+i-1) for deterministic variation
- random seed (-1) batches give independent random seeds per image
- Partial batch failures continue (error TextContent in slot, remaining images proceed)
- Returns flat interleaved [Text1, Image1, Text2, Image2, ...] list
- 34/34 tests passing (was 19, added 15 new tests)
- Use urllib.parse.quote_plus instead of str.replace(' ', '+') for correct
URL encoding of special chars (&, %, +, #, =)
- Add search_url field to return dict so caller can verify/debug the query
- Add result_count field for quick summary without len(results)
- Deduplicate results by URL via seen_urls set
- Filter cards with both empty title AND empty snippet
- Richer hint string: 'Title (url): snippet[:120]' pipe-separated
- Max-results guard now breaks early (no over-fetching)
- 5 new tests (23→28): URL encoding, result_count, dedup, empty filter, hint format
- mcp-builder rules: add wiki/ to structure diagram, add Wiki Update
Workflow section (MANDATORY), update After Building a Server checklist
- gitea-push skill: add wiki deploy as a valid use case
- README.md: add wiki section with deploy_wiki.sh pointer, add
mcp-image-gen to MCP servers table
- Add webscraper_search_hint() tool using Brave Search as backend
(no CAPTCHA/GDPR consent wall, works with plain httpx)
- Add User-Agent header to _fetch_page() — fixes 403 on Wikipedia,
Feynman Lectures, and other sites that block headless requests
- Add 5 new tests for search hint (23 total, 90% coverage)
Brave Search URL: https://search.brave.com/search?q={query}&source=web
Use sparingly — once per research task as orientation, not in loops
The .ach-image div had correct CSS dimensions (64x64) and background-size:cover
but was missing the inline style="background-image: url(...)" — so the div
rendered as an empty circle. Fixed by extracting img_url variable and applying
it as style attribute in the f-string. All 39 achievement PNGs now load.
303/303 tests passing.
- Renamed 19 timestamp-named PNGs (20260404_*) to match original
achievement IDs in profile_builder.py compute_achievements() order:
first_breath, first_thought, eureka, honest_mind, scholar,
deep_knowledge, scientist, veteran, on_fire, storyteller,
night_owl, speed_thinker, first_handshake, birthday, shared_mind,
frugal_mind, quarter_million, token_millionaire, sniper
- Deleted 2 duplicate/excess timestamp PNGs
- Added image= field to all 19 original _add() calls in
profile_builder.py so every achievement now has a PNG path
- All 39 achievements (19 original + 20 tiered) now have image fields
- 303/303 tests pass
FLUX.1-schnell is a gated model — bare wget returns HTTP 401.
- Replace bare wget with huggingface-cli login + download (Option A)
- Add wget with Authorization header as Option B
- Add license acceptance prerequisite (huggingface.co gated repo)
- Add token creation link (huggingface.co/settings/tokens)
- Add fp8 quantized variant as alternative (~8.1GB, faster inference)
- Add download size note (~8GB, 10-30min)
ComfyUI is NOT on PyPI — `pip install comfyui` fails with
"No matching distribution found". Remove the wrong Option A.
Replace with:
- Warning note that pip install does not work
- Only correct method: git clone from GitHub + pip install -r requirements.txt
ROCm status confirmed: rocm-smi 3.1.0 / ROCm-SMI-LIB 7.7.0 installed.
Patrick Plate