Skip to content

V0.04: dashboard overhaul, SSE streaming, episode scoring - #16

Merged
createpjf merged 1 commit into
mainfrom
feat/v0.04-dashboard-streaming-scoring
Feb 26, 2026
Merged

createpjf merged 1 commit into
mainfrom
feat/v0.04-dashboard-streaming-scoring

Conversation

@createpjf

Copy link
Copy Markdown
Owner

Summary

  • Dashboard overhaul: KB horizontal grid with scrollable cards, session sidebar, tab badges with live counts, episode score + timestamp columns, daily log tool artifact filtering
  • SSE streaming: per-task stream files with lockless append, GET /v1/stream/:task_id endpoint, auto-cleanup on task complete
  • Episode scoring chain: baseline scores at creation (success=8, partial=5, other=2) + Alic critique backfill via update_episode_score()
  • New modules: Memo Protocol integration (adapters/memo/), memory consolidation pipeline, TextGrad critique-to-skill feedback, product documentation
  • Other: generate_doc expanded to 8 formats, empty LLM response guards, session history increased to 200, dashboard session API, KB insight dedup

Test plan

  • Start gateway (cleo gateway start) and verify dashboard loads at http://127.0.0.1:19789
  • Verify KB tab shows horizontal card grid with colored tags
  • Create/rename/delete dashboard sessions via sidebar
  • Submit a task and verify SSE streaming shows tokens in real-time
  • Check Episodes tab for score + timestamp columns with data
  • Verify Daily Log filters out tool call artifacts
  • Run pytest tests/ — expect 399 passed

🤖 Generated with Claude Code

Dashboard:
- KB horizontal grid with scrollable cards and tag colors
- Session sidebar (create/rename/delete conversations)
- Tab badges with live counts on Episodes/Cases/KB
- Episode score + timestamp columns
- Daily log content filtering (hide tool artifacts)

SSE Streaming:
- Per-task stream files (lockless append, cursor-based read)
- GET /v1/stream/:task_id endpoint
- Auto-cleanup on task complete/fail/cancel

Episode Scoring:
- Baseline scores at creation (success=8, partial=5, other=2)
- Alic critique backfill via EpisodicMemory.update_episode_score()
- Orchestrator wires critique to evaluated agent's episode

New modules:
- adapters/memo/ — Memo Protocol integration
- adapters/memory/consolidator.py — 3-phase memory consolidation
- reputation/textgrad.py — TextGrad critique → skill patches
- cli/memo_cmd.py — Memo CLI
- docs/Cleo_V0.01_*.md — Product and technical documentation

Other: generate_doc 8 formats, empty LLM guards, session history 200,
dashboard session API, KB insight dedup, cron timeout improvements

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@createpjf
createpjf merged commit b6211b4 into main Feb 26, 2026
3 of 4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant