2025-11-26T13:20:25Z - 2026-05-26T12:20:25Z
Overview
10 pull requests merged by 1 user
Merged
#77 fix(ai): match target root by basename in _apply_plan()
Merged
#75 feat(ai): Phase 3 investigation planning
Merged
#73 chore: update CLAUDE.md for session 9 wrap-up
Merged
#71 test(ai): cover _TokenTracker, _synthesize_from_cache, _discover_directories (#70)
Merged
#69 test(ai): add unit coverage for pure helpers in ai.py (#55)
Merged
#68 refactor(ai): single-source tool registration via register_tool() (#56)
Merged
#67 refactor(ai): extract _run_dir_loop into three focused helpers (#57)
Merged
#66 chore: update CLAUDE.md for session 9
Merged
#65 feat: AI investigation is the product, drop zero-dep constraint (#64)
Merged
#62 Add README and Apache 2.0 LICENSE
24 issues closed from 2 users
Closed
#76 _apply_plan() fails to match target root directory from plan paths
Closed
#8 Add _PLANNING_SYSTEM_PROMPT to prompts.py
Closed
#9 Implement planning pass with submit_plan tool in ai.py
Closed
#10 Implement dynamic turn allocation from plan output
Closed
#11 Save investigation plan to cache for resumed runs
Closed
#74 Add investigation quality instrumentation
Closed
#72 Document the leaf-first investigation contract in Internals.md
Closed
#70 Pre-Phase-3 test coverage: _TokenTracker, _synthesize_from_cache, _discover_directories
Closed
#55 Add unit test coverage for pure helpers in ai.py
Closed
#56 Tool registration is duplicated between _TOOL_DISPATCH and _DIR_TOOLS
Closed
#57 Refactor _run_dir_loop before Phase 3 dynamic turn allocation lands
Closed
#35 Incremental AI re-investigation in watch mode
Closed
#64 Scope change: AI investigation is the product, drop zero-dependency constraint
Closed
#54 Phase 1 confidence-write path is dormant — no prompt instructs the agent to set confidence
Closed
#1 Add confidence fields to file and dir cache schemas
Closed
#53 Developer onboarding: deep-dive code tour + reconcile architecture/roadmap docs
Closed
#44 Dir loop exhausts context budget on small targets
Closed
#42 Rebuild filetype classifier to remove source-code bias
Closed
#7 Skip survey pass for targets below minimum size threshold
Closed
#6 Wire survey output into dir loop system prompt
Closed
#5 Implement _run_survey() and submit_survey tool in ai.py
Closed
#4 Add _SURVEY_SYSTEM_PROMPT to prompts.py
Closed
#3 Add low_confidence_entries() query to CacheManager
Closed
#2 Update dir loop prompt to set confidence on cache writes
66 issues created by 3 users
Opened
#1 Add confidence fields to file and dir cache schemas
Opened
#2 Update dir loop prompt to set confidence on cache writes
Opened
#3 Add low_confidence_entries() query to CacheManager
Opened
#4 Add _SURVEY_SYSTEM_PROMPT to prompts.py
Opened
#5 Implement _run_survey() and submit_survey tool in ai.py
Opened
#6 Wire survey output into dir loop system prompt
Opened
#7 Skip survey pass for targets below minimum size threshold
Opened
#8 Add _PLANNING_SYSTEM_PROMPT to prompts.py
Opened
#9 Implement planning pass with submit_plan tool in ai.py
Opened
#10 Implement dynamic turn allocation from plan output
Opened
#11 Save investigation plan to cache for resumed runs
Opened
#12 Research and decide on web search API backend
Opened
#13 Create luminos_lib/search.py with web_search implementation
Opened
#14 Add package_lookup tool (PyPI, npm, crates.io, pkg.go.dev)
Opened
#15 Add fetch_url tool for URLs referenced in local files
Opened
#16 Add ask_user tool and --interactive flag
Opened
#17 Add --no-external flag and per-session budget tracking
Opened
#18 Implement tier classification (small/medium/large/xlarge)
Opened
#19 Switch small tier synthesis to per-file cache entries
Opened
#20 Add tier-appropriate depth instructions to synthesis prompt
Opened
#21 Implement grouping pass with submit_grouping tool
Opened
#22 Implement two-level grouping for xlarge tier
Opened
#23 Update synthesis prompt for hypothesis formation
Opened
#24 Enable think tool in synthesis pass
Opened
#25 Implement _run_refinement() with --refine flag
Opened
#26 Add --refine-depth N for multiple refinement passes
Opened
#27 Use confidence scores to prioritize refinement targets
Opened
#28 Define superset output fields in synthesis submit_report schema
Opened
#29 Update report formatter for dynamic field rendering
Opened
#30 Add luminos.toml config file for tunable thresholds
Opened
#31 Add structured JSONL investigation log
Opened
#32 Add --security flag for security-focused analysis mode
Opened
#33 Add Markdown and HTML output formats
Opened
#34 Add --model flag and per-pass model configuration
Opened
#35 Incremental AI re-investigation in watch mode
Opened
#36 Add optional PDF and Office document readers
Opened
#37 Add unit test coverage for existing modules
Opened
#38 Add cache invalidation based on file mtime
Opened
#39 Phase 3.5: Migrate to MCP backend architecture
Opened
#40 Review and update Phase 4+ issues after MCP pivot
Opened
#42 Rebuild filetype classifier to remove source-code bias
Opened
#44 Dir loop exhausts context budget on small targets
Opened
#46 Revisit survey-skip thresholds with empirical data
Opened
#48 Unit of analysis is hardcoded as "file" — over/under-counts container formats
Opened
#49 Terminal report shows biased bucketed file-type view
Opened
#51 Dir loop message history grows monotonically — no eviction
Opened
#53 Developer onboarding: deep-dive code tour + reconcile architecture/roadmap docs
Opened
#54 Phase 1 confidence-write path is dormant — no prompt instructs the agent to set confidence
Opened
#55 Add unit test coverage for pure helpers in ai.py
Opened
#56 Tool registration is duplicated between _TOOL_DISPATCH and _DIR_TOOLS
Opened
#57 Refactor _run_dir_loop before Phase 3 dynamic turn allocation lands
Opened
#63 Automate Forgejo-to-GitHub wiki mirror sync
Opened
#64 Scope change: AI investigation is the product, drop zero-dependency constraint
Opened
#70 Pre-Phase-3 test coverage: _TokenTracker, _synthesize_from_cache, _discover_directories
Opened
#72 Document the leaf-first investigation contract in Internals.md
Opened
#74 Add investigation quality instrumentation
Opened
#76 _apply_plan() fails to match target root directory from plan paths
Opened
#78 Persist AI synthesis output to cache and support file output
Opened
#79 Stale cache entries survive --fresh when path format changed
Opened
#80 security(ai): run_command uses shell=True, whitelist is bypassable via shell metachars
Opened
#81 fix(ai): magic.from_file() raises MagicException, only OSError is caught — can crash dir loop
Opened
#82 fix(ast_parser): parse_structure has no exception handling around parser/handler — crashes dir loop on malformed input
Opened
#83 fix(ai): _run_dir_loop APIError path doesn't flush partial dir entry, loses cached work
Opened
#84 fix(cache): write_entry missing validation for Phase 3 completeness field
Opened
#85 fix(cache): _save_investigations write is not atomic — crash mid-write corrupts shared UUID map
Opened
#86 fix(recency): find -printf uses \n record separator — breaks on filenames containing newlines