Skip to content

feat(kb): Matrix chat transcript → KB claims extraction pipeline - #56

Merged
mdheller merged 1 commit into
mainfrom
feat/transcript-kb-extract
Aug 4, 2026
Merged

feat(kb): Matrix chat transcript → KB claims extraction pipeline#56
mdheller merged 1 commit into
mainfrom
feat/transcript-kb-extract

Conversation

@mdheller

@mdheller mdheller commented Aug 4, 2026

Copy link
Copy Markdown

Summary

  • Adds turtle-transcript-extract binary: fetches Matrix room history, groups messages into substantive threads, distills each via turtle-noetica-stream, writes KB claims to ~/.local/state/sourceos/transcript-claims/pending.jsonl, and commits them as YAML to systems-learning-loops/kb/claims/
  • Adds txextract, txcommit, txlist shell functions to turtle-shell-init.zsh
  • Adds three palette entries (📜 Transcript) to turtleterm.lua with matching dispatch actions

Claim format

Committed claims are YAML with id, extracted_at, source, room_id, thread_event_ids, tags, claim, confidence, evidence_summary.

Thread detection heuristics

A thread is substantive if: ≥3 messages, or question (?) + reply, or !qes operator command + reply, or ≥2 messages with any body >100 chars.

Test plan

  • python3 -m py_compile assets/sourceos/bin/turtle-transcript-extract — passes
  • turtle-transcript-extract status — shows config/token/pending state
  • turtle-transcript-extract extract --dry-run — extracts without writing
  • txextract --since 1 in TurtleTerm shell — invokes binary via shell alias
  • Palette CMD+P → search "transcript" → entries appear, SendString fires correctly

@mdheller
mdheller force-pushed the feat/transcript-kb-extract branch from 20ba18d to 887c0c1 Compare August 4, 2026 15:52
@mdheller
mdheller merged commit 78df9c9 into main Aug 4, 2026
14 checks passed
@mdheller
mdheller deleted the feat/transcript-kb-extract branch August 4, 2026 15:52
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