my-pi 0.1.7 → 0.1.9

package/README.md

@@ -7,9 +7,8 @@ Composable [pi](https://pi.dev) coding agent for humans and agents.
 
 Built on the
 [@mariozechner/pi-coding-agent](https://github.com/badlogic/pi-mono)
-SDK. Adds MCP server support, extension stacking, LSP tools, agent
-chains, prompt presets, local SQLite telemetry for evals, and a
-programmatic API.
+SDK. Adds MCP server support, extension stacking, LSP tools, prompt
+presets, local SQLite telemetry for evals, and a programmatic API.
 
 Extension stacking patterns inspired by
 [pi-vs-claude-code](https://github.com/disler/pi-vs-claude-code).
@@ -22,16 +21,14 @@ Extension stacking patterns inspired by
   `mcp.json`, auto-registered as Pi tools.
 - **Built-in LSP tools** — diagnostics, hover, definitions,
   references, and document symbols via language servers.
-- **Agent chains** — sequential scout/plan style workflows defined in
-  `.pi/agents/agent-chain.yaml`.
 - **Managed skills** — discover, enable, disable, import, and sync
   Pi-native skills.
 - **Prompt presets** — base presets plus additive prompt layers with
   per-project persistence.
 - **Secret redaction** — redact API keys and other sensitive output
   before the model sees tool results.
-- **Session handoff + recall** — export focused handoff markdown and
-  teach the model to use `pirecall` for prior-session context.
+- **Recall** — teach the model to use `pirecall` for prior-session
+  context.
 - **Local telemetry** — optional SQLite telemetry for evals, tool
   analysis, and operational debugging.
 - **Bundled themes + extension stacking** — ship defaults, then layer
@@ -85,10 +82,9 @@ Outputs NDJSON events — one JSON object per line — for programmatic
 consumption by other agents or scripts.
 
 In non-interactive modes (`"prompt"`, `-P`, `--json`), my-pi keeps
-headless-capable built-ins like MCP, LSP, chains, prompt presets,
-recall, hooks, and secret filtering enabled, while skipping UI-only
-built-ins like handoff, confirm-destructive, session auto-naming, and
-working-indicator customization.
+headless-capable built-ins like MCP, LSP, prompt presets, recall,
+hooks, and secret filtering enabled, while skipping UI-only built-ins
+like session auto-naming.
 
 ### Local telemetry (SQLite)
 
@@ -160,7 +156,7 @@ Query and export helpers:
 
 Schema notes:
 
-- source of truth: `src/extensions/telemetry-schema.sql`
+- source of truth: `packages/pi-telemetry/src/schema.sql`
 - current telemetry schema version: `1`
 - schema version is tracked with `PRAGMA user_version`
 - unversioned local telemetry databases are initialized/upgraded to v1
@@ -169,7 +165,6 @@ Schema notes:
   downgrading
 - opens the database in WAL mode: `PRAGMA journal_mode = WAL`
 - waits up to 5s on lock contention: `PRAGMA busy_timeout = 5000`
-- packaged builds ship the schema as `dist/telemetry-schema.sql`
 
 CLI flags `--telemetry` and `--no-telemetry` override only the current
 process. `/telemetry on` and `/telemetry off` update the saved default
@@ -342,24 +337,29 @@ In interactive mode:
 - `/skills sync <key|name>` — sync an imported skill to its upstream
 - `/skills refresh` — rescan skill directories
 - `/skills defaults <all-enabled|all-disabled>` — set default policy
-- `/preset` — open the prompt preset manager (base presets + layers)
-- `/preset <name>` — activate a base preset or toggle a layer
-- `/preset base <name>` — activate a base preset directly
-- `/preset enable <layer>` / `/preset disable <layer>` — toggle a
-  prompt layer directly
-- `/preset edit <name>` — edit or create a project-local preset in
-  `.pi/presets.json`
-- `/preset delete <name>` — delete a project-local preset
-- `/preset reset <name>` — remove a project-local override and fall
-  back to user/built-in if available
-- `/preset clear` — clear the active base preset and all layers
-- `/chain` — inspect or switch the active agent chain
-- `/agents` — list discovered agent definitions
+- `/prompt-preset` — open the prompt preset manager (base presets +
+  layers); `/preset` is a short alias
+- `/prompt-preset help` — show examples and common prompt preset
+  commands
+- `/prompt-preset <name>` — activate a base preset or toggle a layer
+- `/prompt-preset base <name>` — activate a base preset directly
+- `/prompt-preset enable <layer>` / `/prompt-preset disable <layer>` —
+  toggle a prompt layer directly
+- `/prompt-preset edit <name>` — edit or create a project preset in
+  `.pi/presets/<name>.md`
+- `/prompt-preset edit-global <name>` — edit or create a global preset
+  in `~/.pi/agent/presets/<name>.md`
+- `/prompt-preset export-defaults` — copy built-in presets to editable
+  global Markdown files
+- `/prompt-preset export-defaults project` — copy built-in presets to
+  editable project Markdown files
+- `/prompt-preset delete <name>` — delete a project-local preset
+- `/prompt-preset reset <name>` — remove a project-local override and
+  fall back to user/built-in if available
+- `/prompt-preset clear` — clear the active base preset and all layers
 - `/lsp status|list|restart` — inspect or restart language server
   state
 - `/redact-stats` — show how many secrets were redacted this session
-- `/handoff <task>` — export current context to a handoff markdown
-  file
 - `/telemetry status|stats|query|export|on|off|path` — inspect, query,
   export, or toggle local SQLite telemetry
 
@@ -376,36 +376,6 @@ In interactive mode:
    persisted in `~/.config/my-pi/extensions.json`
 8. Cleanup on `session_shutdown`
 
-## Agent Chains
-
-Define sequential agent pipelines in `.pi/agents/agent-chain.yaml`:
-
-```yaml
-scout-plan:
-  description: 'Scout the codebase then plan implementation'
-  steps:
-    - agent: scout
-      prompt: 'Explore and analyze: $INPUT'
-    - agent: planner
-      prompt: 'Based on this analysis, create a plan:\n\n$INPUT'
-```
-
-Agent definitions live in `.pi/agents/*.md` with frontmatter:
-
-```markdown
----
-name: scout
-description: Codebase exploration and analysis
-tools: read,grep,find,ls
----
-
-You are a scout agent. Explore the codebase and report findings.
-```
-
-The chain extension injects context into the system prompt so the LLM
-knows when and how to use `run_chain`. Use `/chain` to switch active
-chains and `/agents` to list available agents.
-
 ## Secret Redaction
 
 The filter-output extension automatically redacts secrets (API keys,
@@ -441,11 +411,30 @@ Preset sources are merged in this order:
 
 1. built-in defaults
 2. `~/.pi/agent/presets.json`
-3. `.pi/presets.json`
+3. `~/.pi/agent/presets/*.md`
+4. `.pi/presets.json`
+5. `.pi/presets/*.md`
 
 Project presets override global/default presets with the same name.
-Strings are treated as base presets by default. Object entries may set
-`kind: "base"` or `kind: "layer"`.
+Strings in JSON are treated as base presets by default. Object entries
+may set `kind: "base"` or `kind: "layer"`. Markdown preset files use
+the filename as the preset name and optional frontmatter:
+
+```markdown
+---
+kind: base
+description: Short, direct, no fluff
+---
+
+Be concise and direct.
+```
+
+Use `/prompt-preset export-defaults` to copy built-in presets to
+`~/.pi/agent/presets/*.md` for editing, or
+`/prompt-preset export-defaults project` to write `.pi/presets/*.md`.
+`/prompt-preset edit <name>` writes a project Markdown preset;
+`/prompt-preset edit-global <name>` writes a global one. `/preset` is
+a short alias for `/prompt-preset`.
 
 CLI layering is supported too:
 
@@ -502,54 +491,72 @@ clients.
 
 ## Session Recall
 
-The recall extension nudges the model to use `npx pirecall` when the
-user references prior work or when historical project context would
-help. It also triggers a background `pirecall sync --json` on session
-start when the local recall database exists.
+The recall package nudges the model to use `pnpx pirecall` or
+`npx pirecall` when the user references prior work or when historical
+project context would help. It also triggers `pirecall sync --json` on
+session start and shutdown when the local recall database exists.
 
-## Session Handoff
+## Reusable Pi packages
 
-Use `/handoff <task>` to export conversation context as a markdown
-file that can be piped into a new session:
+This repo is a pnpm workspace. The `my-pi` harness depends on reusable
+Pi packages via `workspace:*`, and those packages can also be
+published and installed into vanilla `pi` independently:
 
 ```bash
-# In session 1: /handoff continue the auth refactor
-# Then:
-pnpx my-pi@latest < handoff-1234567890.md
+pi install npm:@spences10/pi-redact
+pi install npm:@spences10/pi-telemetry
+pi install npm:@spences10/pi-mcp
+pi install npm:@spences10/pi-lsp
+pi install npm:@spences10/pi-confirm-destructive
+pi install npm:@spences10/pi-skills
+pi install npm:@spences10/pi-recall
+pi install npm:@spences10/pi-nopeek
 ```
 
+- [`@spences10/pi-redact`](./packages/pi-redact/README.md) — output
+  redaction and `/redact-stats`
+- [`@spences10/pi-telemetry`](./packages/pi-telemetry/README.md) —
+  local SQLite telemetry and `/telemetry`
+- [`@spences10/pi-mcp`](./packages/pi-mcp/README.md) — MCP server
+  integration and `/mcp`
+- [`@spences10/pi-lsp`](./packages/pi-lsp/README.md) — LSP-backed
+  diagnostics and symbol tools
+- [`@spences10/pi-confirm-destructive`](./packages/pi-confirm-destructive/README.md)
+  — destructive action confirmations
+- [`@spences10/pi-skills`](./packages/pi-skills/README.md) — skill
+  management, import, and sync
+- [`@spences10/pi-recall`](./packages/pi-recall/README.md) — pirecall
+  reminder and background sync
+- [`@spences10/pi-nopeek`](./packages/pi-nopeek/README.md) — nopeek
+  reminder for secret-safe environment loading
+
+Each package README is the entry point for install instructions,
+commands, runtime behavior, and development notes.
+
 ## Project Structure
 
 ```
 src/
-  index.ts            CLI entry point (citty + pi SDK)
-  api.ts              Programmatic API (create_my_pi + re-exports)
+  index.ts                 CLI entry point (citty + pi SDK)
+  api.ts                   Programmatic API (create_my_pi + re-exports)
   extensions/
-    config.ts         Persistent built-in extension config
-    extensions.ts     Built-in extension manager (/extensions)
-    mcp.ts            MCP server integration
-    skills.ts         Skill discovery and toggle
-    chain.ts          Agent chain pipelines
-    filter-output.ts  Secret redaction in tool output
-    handoff.ts        Session context export
-    lsp.ts            Language server tools and /lsp command
-    telemetry.ts      Local telemetry extension and /telemetry command
-    prompt-presets.ts Runtime prompt preset selection and editing
-    recall.ts         Past session recall guidance
-    telemetry-*.ts    Telemetry config + SQLite storage
-  mcp/
-    client.ts         Minimal MCP stdio client (JSON-RPC 2.0)
-    config.ts         Loads and merges mcp.json configs
-  skills/
-    manager.ts        Skill enable/disable state management
-    scanner.ts        Skill discovery across sources
-    config.ts         Persistent skills config (~/.config/my-pi/)
+    manager/               Built-in extension manager and config
+    prompt-presets/        Runtime prompt preset selection and editing
+    session-name/          Session auto-naming
+    hooks-resolution/      Claude-style hook resolution
+packages/
+  pi-redact/               Installable Pi package for output redaction
+  pi-telemetry/            Installable Pi package for SQLite telemetry
+  pi-mcp/                  Installable Pi package for MCP integration
+  pi-lsp/                  Installable Pi package for LSP tools
+  pi-confirm-destructive/  Installable Pi package for destructive action confirmations
+  pi-skills/               Installable Pi package for skill management
+  pi-recall/               Installable Pi package for pirecall reminders
+  pi-nopeek/               Installable Pi package for nopeek reminders
 .pi/
-  agents/
-    *.md              Agent definitions (frontmatter + system prompt)
-    agent-chain.yaml  Chain pipeline definitions
-  presets.json        Optional project prompt presets
-mcp.json              Project MCP server config
+  presets.json             Optional project prompt presets (JSON)
+  presets/*.md             Optional project prompt presets (Markdown files)
+mcp.json                   Project MCP server config
 ```
 
 ## Development