teleo/teleo-infrastructure

Fork 0

Commit graph

Author	SHA1	Message	Date
m3taversal	c0a6adf9ed	leo: model diversity + calibrated review prompts - Domain review → GPT-4o (OpenRouter), Leo STANDARD → Sonnet (OpenRouter), Leo DEEP → Opus (Claude Max). Two model families = no correlated blind spots. - Opus reserved for DEEP eval only — protects rate limit for overnight research. - Review prompts calibrated: require per-criterion evidence, blocking-vs-observation verdict rules. Moved from 100% rubber-stamp approval to 12% pass rate. - OpenRouter failures classified as openrouter_failed (not rate_limited) to avoid spurious 15-min Opus backoff. - merge.py: pre-check PR state before merge API call (prevents 405 on re-merge). Pentagon-Agent: Leo <294C3CA1-0205-4668-82FA-B984D54F48AD>	2026-03-13 17:10:30 +00:00
m3taversal	85b86a918a	ganymede: extract lib/llm.py from evaluate.py (Phase 3c) Some checks failed CI / lint-and-test (pull_request) Has been cancelled Details - What: LLM transport (OpenRouter, Claude CLI), prompt templates (triage/domain/Leo), and review runner functions moved to lib/llm.py. evaluate.py retains PR lifecycle orchestration, SQLite state, Forgejo posting, rate limit backoff, and evaluate_cycle. - Why: evaluate.py was 734 lines mixing orchestration with LLM concerns. Now 455 lines orchestration + 250 lines LLM transport. Each module has a single responsibility. - Connections: completes Phase 3 structural refactor (forgejo.py + domains.py + llm.py). teleo-pipeline.py updated to import kill_active_subprocesses from lib.llm. Pentagon-Agent: Ganymede <F99EBFA6-547B-4096-BEEA-1D59C3E4028A>	2026-03-13 15:40:18 +00:00

Author

SHA1

Message

Date

m3taversal

c0a6adf9ed

leo: model diversity + calibrated review prompts

- Domain review → GPT-4o (OpenRouter), Leo STANDARD → Sonnet (OpenRouter),
  Leo DEEP → Opus (Claude Max). Two model families = no correlated blind spots.
- Opus reserved for DEEP eval only — protects rate limit for overnight research.
- Review prompts calibrated: require per-criterion evidence, blocking-vs-observation
  verdict rules. Moved from 100% rubber-stamp approval to 12% pass rate.
- OpenRouter failures classified as openrouter_failed (not rate_limited) to avoid
  spurious 15-min Opus backoff.
- merge.py: pre-check PR state before merge API call (prevents 405 on re-merge).

Pentagon-Agent: Leo <294C3CA1-0205-4668-82FA-B984D54F48AD>

2026-03-13 17:10:30 +00:00

m3taversal

85b86a918a

ganymede: extract lib/llm.py from evaluate.py (Phase 3c)

CI / lint-and-test (pull_request) Has been cancelled

Details

- What: LLM transport (OpenRouter, Claude CLI), prompt templates
  (triage/domain/Leo), and review runner functions moved to lib/llm.py.
  evaluate.py retains PR lifecycle orchestration, SQLite state, Forgejo
  posting, rate limit backoff, and evaluate_cycle.
- Why: evaluate.py was 734 lines mixing orchestration with LLM concerns.
  Now 455 lines orchestration + 250 lines LLM transport. Each module has
  a single responsibility.
- Connections: completes Phase 3 structural refactor (forgejo.py + domains.py
  + llm.py). teleo-pipeline.py updated to import kill_active_subprocesses
  from lib.llm.

Pentagon-Agent: Ganymede <F99EBFA6-547B-4096-BEEA-1D59C3E4028A>

2026-03-13 15:40:18 +00:00

2 commits