v0.1.1 — fix observe capture, add CRUD tools, honest skill gating - #3
Merged
Conversation
observe() ran only first-person-anchored rules (^i prefer, ^we decided) and silently discarded everything else — third-person statements, emails, most natural prose. A local user feeding raw text saw nothing persist, which defeats the core 'feed it text and it remembers' UX. Add a fallback in extract(): when no structured rule matches but the input is substantive prose (not a question, not code, not an ack), persist it verbatim as a raw 'observation' memory. Structured rules still win when they match; hosted LLM extraction remains the premium that refines these into typed facts. Fixes the smoke test: 1/4 -> 4/4 statements captured. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The OSS MCP server exposed 7 tools; the plugin's forget/dream/stats/health skills call memory_delete / memory_supersede / memory_stats, which only existed on the proprietary superset — so those skills errored on a local install. These three are basic hygiene, not moat: the Storage trait already implements delete/supersede/count_items. Wire them into the MCP surface. OSS MCP surface: 7 -> 10 tools. Prediction / graph-reasoning / build-context remain proprietary-only (the actual moat). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The plugin shipped predict/behaviors/why/context-loader/graph skills that call MCP tools only the hosted engine registers (memory_predict, build_context, graph_reason, query_graph, prefetch_related). On a local/OSS install they'd error. Add a 'requires hosted mode' banner to each so the agent degrades to search/recall instead of calling a missing tool; graph notes that BUILDING the graph works locally, only reasoning needs hosted. Retune plugin + marketplace descriptions to lead with the local-first memory store (the OSS value) and frame prediction/behavior as the hosted upgrade, rather than advertising moat features a local install can't run. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
First patch release. Fixes a silent data-loss bug in
observeand aligns the plugin with what the OSS engine actually does.Fixes
observeno longer drops substantive input — the heuristic extractor only matched first-person rules (^I prefer,^we decided) and silently discarded everything else (third-person statements, emails, most prose). Now, when no structured rule matches, substantive text is persisted verbatim as a rawobservation. Smoke test: 1/4 → 4/4 statements captured. Questions/filler/code still filtered. (fix/observe-raw-persist)memory_delete,memory_supersede,memory_stats(OSS MCP surface 7 → 10). Un-breaks theforget/dream/stats/healthskills that called them. Prediction / graph-reasoning / build-context remain hosted-only (the moat).predict/behaviors/why/context-loader/graphskills now carry a "requires hosted mode" banner so a local install degrades tosearch/recallinstead of calling a missing tool. Plugin + marketplace descriptions retuned to lead with the local-first memory store.Verification
tools/listover real stdio returns all 10 toolsobserve→searchroundtrip confirmed on a fresh DB🤖 Generated with Claude Code