chat-recall 0.0.1 → 0.4.8

package/skills/chat-recall-memory/SKILL.md

@@ -0,0 +1,55 @@
+---
+name: chat-recall-memory
+description: >-
+  Store and recall durable facts, decisions, and notes with chat-recall's
+  knowledge graph and diary. Use when the user says "remember this", "note that
+  we decided…", "what do you know about X", "did we decide Y", "what's our
+  convention for Z", or when a non-obvious decision/preference is made that future
+  sessions should know. Covers the temporal knowledge graph, decision records,
+  the agent diary, and a scoped key/value store.
+---
+
+# Durable memory with chat-recall
+
+Two directions: **recall** what's already known, and **record** new facts so
+future sessions inherit them.
+
+## Recall what's known
+
+- `mcp__chat-recall__recall_kg_query` — query an entity's relationships (facts
+  with validity windows). `as_of: <date>` time-travels to what was true then.
+  Use to VERIFY before asserting a fact ("did we decide to use Postgres?").
+- `mcp__chat-recall__recall_kg_timeline` — chronological story of an entity (or
+  everything). `mcp__chat-recall__recall_kg_stats` — graph overview.
+- `mcp__chat-recall__recall_wake_up` — high-importance facts + KG snapshot for a
+  fast "what's true right now".
+- `mcp__chat-recall__recall_diary_read` — prior agent diary entries (narrative
+  notes across sessions). `mcp__chat-recall__recall_get` — a scoped key/value.
+
+## Record new knowledge (do this when something non-obvious is decided)
+
+- `mcp__chat-recall__recall_decision_record` — an explicit decision (subject,
+  decision, reason, importance). Shows up in future wake-up context. Use the
+  moment you and the user agree on something a later session would need.
+- `mcp__chat-recall__recall_kg_add` — assert a fact triple
+  (subject → predicate → object, optional `valid_from`). E.g.
+  `("chat-recall", "uses", "Postgres")`.
+- `mcp__chat-recall__recall_kg_invalidate` — mark a fact no longer true when a
+  decision is reversed or a tool is replaced (keeps the graph honest).
+- `mcp__chat-recall__recall_diary_write` — narrative notes ("what I learned /
+  what matters") keyed per agent, across sessions.
+- `mcp__chat-recall__recall_set` — small structured state (current PR url,
+  branch, preferred test runner); `scope` namespaces per project.
+
+## When to record
+
+Record when a decision or preference is (a) non-obvious, (b) not already captured
+by the code/CLAUDE.md/git history, and (c) useful to a future session. Prefer
+`recall_decision_record` for choices with a rationale, `recall_kg_add` for plain
+entity-relationship facts, `recall_diary_write` for narrative context, `recall_set`
+for transient pointers. Don't duplicate what the repo already records.
+
+## Verify-then-assert
+
+Before stating "we decided X" or "we use Y", confirm with `recall_kg_query`
+(and `as_of` if the timing matters) rather than relying on memory.

package/skills/chat-recall-project/SKILL.md

@@ -0,0 +1,75 @@
+---
+name: chat-recall-project
+description: >-
+  Get the state of a project from chat-recall. Use when the user asks "what's
+  been happening in <project>", "catch me up on this repo", "what's the state of
+  X", about time/cost/activity across a project, or about code findings and
+  recommended fixes. Also owns the shared task board and the team view — "what
+  are my open tasks", "create a task for…", "what is my team working on", "which
+  projects are shared". Aggregates sessions, plans, tasks, commits, cost,
+  knowledge graph, and code-intelligence for one project.
+---
+
+# Project intelligence with chat-recall
+
+## The one-call overview
+
+`mcp__chat-recall__recall_project_context` — the full project dossier: overview,
+tech stack, decisions log, recent sessions with summaries, open tasks, plans,
+diary conclusions, and cost. Accepts a path, a name substring, or a project id
+(`git:host/owner/repo`, `git-local:<sha1>` for a repo with no remote, `ws:name`,
+`user:custom`). Start here for "catch me up on this project".
+
+## Activity, cost, trends
+
+- `mcp__chat-recall__recall_weekly_digest` — sessions, cost, top projects, trend
+  vs last week, open tasks, git activity. `weeks_back: 0` = this week.
+- `mcp__chat-recall__recall_analytics_summary` — cross-tool spend & activity
+  totals (sessions, cost, top tools/projects/models). This one is NOT
+  project-scoped, so it is the right call for "how much did I spend this month"
+  across everything.
+- `mcp__chat-recall__recall_outcome_summary` — how sessions ended over the last
+  N days (shipped / interrupted / abandoned), for "am I finishing what I start".
+
+## Code intelligence (when the codeindex companion is present)
+
+- `mcp__chat-recall__recall_code_projects` — indexed projects with health score.
+- `mcp__chat-recall__recall_code_findings` — findings (security / clones / dead
+  code / coupling / cycles …); each carries a ready-to-run agent prompt. Filter
+  by `project`, `severity`, `category`.
+- `mcp__chat-recall__recall_code_actions` — the ranked "what to fix next" plan.
+- `mcp__chat-recall__recall_recommendations` — concrete recommendations
+  (`scope: 'project'` with a `project` id) reasoning over code + how sessions went.
+- If a repo isn't indexed yet, `mcp__chat-recall__recall_code_index` (path) runs
+  the analyzer and syncs findings.
+
+## Tasks and team (shared, not personal notes)
+
+- `mcp__chat-recall__recall_tasks` — the shared task board: status, assignee,
+  project. This is the answer to "what are my open tasks".
+- `mcp__chat-recall__recall_task_create` — create a task, optionally assigned to
+  a teammate and attached to a project. Returns the new task id.
+- `mcp__chat-recall__recall_task_update` — change status, reassign, rename or
+  comment. Takes the `t_…` id from `recall_tasks`.
+- `mcp__chat-recall__recall_team_activity` — per-teammate × per-project rollup
+  (session counts, last activity) for "what is the team working on".
+- `mcp__chat-recall__recall_shares` — which projects are shared into the team
+  (`scope: 'mine'` / `'all'`). Sharing is opt-in: nothing is visible to
+  teammates until it is shared, so an empty result means "not shared", NOT
+  "no work happened".
+
+Creating or updating a task is a WRITE other people see. Do it when the user
+asks, not to tidy up on their behalf.
+
+## Workflow
+
+1. "Catch me up on <project>" → `recall_project_context`.
+2. "How productive / how much did X cost" → `recall_weekly_digest` (per week) or
+   `recall_analytics_summary` (cross-project totals).
+3. "What should I fix in <project>" → `recall_code_actions` for the ranked plan,
+   `recall_code_findings` for the raw findings, `recall_recommendations` when the
+   user wants reasoning over how sessions actually went, not just code shape.
+   Index first with `recall_code_index` if the repo is not indexed yet.
+4. "What are my/our open tasks" → `recall_tasks`; "what is the team doing" →
+   `recall_team_activity`.
+5. Ground answers in what you found (session ids, finding ids); don't invent status.

package/skills/chat-recall-resume/SKILL.md

@@ -0,0 +1,62 @@
+---
+name: chat-recall-resume
+description: >-
+  Resume or cold-start work using chat-recall. Use when the user says "continue",
+  "pick up where we left off", "resume our work", "carry on with X", starts a new
+  session on an ongoing task, or references something from "last time" that isn't
+  in the current context. Rebuilds prior context from past sessions instead of
+  asking the user to re-explain.
+---
+
+# Resuming work with chat-recall
+
+Goal: reconstruct the state of prior work so you can continue seamlessly.
+
+## Default move
+
+`mcp__chat-recall__recall_smart_resume` returns a structured bundle: what was
+done, what's pending, files touched, budget, and current knowledge-graph facts.
+This is the best single call for "continue".
+
+**It REQUIRES `session_id` — there is no "latest session" default, and calling
+it bare is a validation error, not a lucky guess.** So resuming is always two
+steps unless the user handed you an id:
+
+1. Get the id. `mcp__chat-recall__recall_recent` (`project_filter`,
+   `since_hours`) lists candidates newest-first — the right call for a plain
+   "continue". Use `mcp__chat-recall__recall_search` with
+   `include_outcome: true` instead when the user described the TASK rather than
+   the timing; the outcome line (shipped / interrupted / abandoned) tells you
+   whether that session is even worth resuming.
+2. Pass that id to `recall_smart_resume`.
+
+`recall_smart_resume` also takes no project filter. Scope the SEARCH in step 1
+(`project_filter` on `recall_recent`), not the resume call.
+
+## Cold start (new session, unclear task)
+
+- `mcp__chat-recall__recall_wake_up` — identity blurb + high-importance
+  decisions/preferences/milestones + a knowledge-graph snapshot. Fast "who am I,
+  what's true right now". Add a project filter to scope it.
+- `mcp__chat-recall__recall_project_context` — when the user is clearly working in
+  one project and wants its full picture (see `chat-recall-project`).
+
+## Reading a specific past session
+
+Once you have a session id:
+- `mcp__chat-recall__recall_context` — structured context (requests, decisions,
+  files, tools).
+- `mcp__chat-recall__recall_show` — raw slice; `from_end: N` for the last N
+  messages, `include_code: true` when the actual diffs/commands matter. It
+  returns 10 messages unless you raise `max_messages`, so don't read a truncated
+  slice as the whole session.
+- `mcp__chat-recall__recall_summary` — AI summary + outcome (did the work land?).
+
+## Workflow
+
+1. Have a session id? → `recall_smart_resume` (or `recall_context`/`recall_show`).
+2. No id, know the task? → `recall_search` (`include_outcome: true`) → take the id from the best hit → `recall_smart_resume`.
+3. No id, no clear task? → `recall_wake_up` (+ project filter), then confirm direction with the user.
+4. Continue the work; state which past session you picked up so the user can course-correct.
+
+Don't ask the user to recap what a recall call can reconstruct.

package/README.md

@@ -1,7 +0,0 @@
-# Reserved
-
-Cross-tool memory for AI coding sessions — Claude Code, Gemini CLI,
-Codex, OpenCode and Antigravity.
-
-This version is a name reservation. The first working release lands
-shortly. See https://chatrecall.dev