refactor(memory): streamline skill for pi-web users, add project/session context

.pi/skills/memory/README.md

@@ -1,16 +1,22 @@
 # Memory Skill
 
-Project-local memory workflow for Set Kyar's workspace.
+Project-local memory workflow for pi-web users.
 
 ## Files
 - `SKILL.md` — skill instructions
 - `scripts/memory.py` — CLI implementation
-- `../../../data/memory.sqlite` — database
-- `../../../data/schema.sql` — schema
+- `$PI_CODING_AGENT_DIR/pi-web-memory.sqlite` (default `~/.pi/agent/`) — database, auto-initialized on first use
+- `data/schema.sql` — schema (shipped with the skill)
 
 ## Use from repo root
 ```bash
-python3 .pi/skills/memory/scripts/memory.py search "rent"
-python3 .pi/skills/memory/scripts/memory.py add-memory "Prefers concise answers" --category preference --importance 4
-python3 .pi/skills/memory/scripts/memory.py add-reminder "Pay rental" --due-at "2026-05-21 09:00" --timezone Asia/Singapore
+# Optional: initialize explicitly (normal commands also auto-initialize)
+python3 .pi/skills/memory/scripts/memory.py init
+
+# Remember a project fact (always scoped to cwd/project)
+python3 .pi/skills/memory/scripts/memory.py add-memory "Prefers tabs over spaces" \
+  --category preference --importance 4 --cwd /path/to/project --project my-project
+
+# Search memories for the current project
+python3 .pi/skills/memory/scripts/memory.py search "tabs" --project my-project
 ```

.pi/skills/memory/SKILL.md

@@ -1,35 +1,75 @@
 ---
 name: memory
-description: Manages project-local long-term memory in SQLite for Set Kyar's workspace. Use when storing, searching, correcting, or listing memories, reminders, accounts, automations, or schema changes.
+description: Manages project-local long-term memory in SQLite for pi-web users. Use when storing, searching, correcting, or listing memories.
 ---
 
 # Memory
 
-Use this skill for project memory tasks.
+Use this skill for project memory tasks. Every memory is scoped to a specific project (working directory) so facts stay relevant to the right context.
 
 ## Core files
-- `data/memory.sqlite` — primary memory database
-- `data/schema.sql` — schema definition
+- `$PI_CODING_AGENT_DIR/pi-web-memory.sqlite` (default `~/.pi/agent/`) — primary memory database, auto-initialized on first use
+- `.pi/skills/memory/data/schema.sql` — schema definition (shipped with the skill)
 - `.pi/skills/memory/scripts/memory.py` — CLI implementation
 
+## Project context
+
+Memories are associated with a project via `--cwd` and `--project`. This allows the same database to serve multiple projects without cross-polluting recall. When a pi session is active, also capture session info so you can trace which conversation produced a memory.
+
+| Context field | Source | Example |
+|---|---|---|
+| `--cwd` | `pwd` / pi session header `cwd` | `/Users/setkyar/pi-web` |
+| `--project` | basename of cwd | `pi-web` |
+| `--session-id` | pi session header `id` (8-char UUID) | `a1b2c3d4` |
+| `--session-name` | `/name` or `session_info` entry | `Refactor auth module` |
+
+Pi sessions are stored as JSONL trees in `getSessionsDir()` (default `~/.pi/agent/sessions/<encoded-cwd>/`). Each session file has a header with `id`, `cwd`, `timestamp`, and optional `parentSession`. Entries form a tree via `id`/`parentId`; use `/tree` to navigate branches, `/fork` to split into a new session, and `/name` to set a display name.
+
+When the user says "remember that" or "save this for later", record:
+- The fact itself (`add-memory`)
+- `--cwd` of the current project
+- `--project` (derived from cwd basename)
+- Current pi session ID and name if a pi session is active
+
 ## Common commands
 ```bash
-python3 .pi/skills/memory/scripts/memory.py search "rent"
-python3 .pi/skills/memory/scripts/memory.py add-memory "Prefers concise answers" --category preference --importance 4
-python3 .pi/skills/memory/scripts/memory.py add-reminder "Pay rental" --due-at "2026-05-21 09:00" --timezone Asia/Singapore
-python3 .pi/skills/memory/scripts/memory.py add-account "Syfe" --account-name "Brokerage" --currency SGD --balance 100000
-python3 .pi/skills/memory/scripts/memory.py schema-changes
+# Optional: initialize explicitly (normal commands also auto-initialize)
+python3 .pi/skills/memory/scripts/memory.py init
+
+# Remember a project fact
+python3 .pi/skills/memory/scripts/memory.py add-memory "Prefers tabs over spaces" \
+  --category preference --importance 4 --cwd /Users/setkyar/pi-web --project pi-web
+
+# Remember with session context
+python3 .pi/skills/memory/scripts/memory.py add-memory "Decided to use Go 1.25 for the backend" \
+  --category decision --importance 5 --cwd /Users/setkyar/pi-web --project pi-web \
+  --session-id a1b2c3d4 --session-name "pi-web backend refactor"
+
+# Search across all projects or filter by project
+python3 .pi/skills/memory/scripts/memory.py search "tabs" --project pi-web
 ```
 
 ## Rules
 - Store explicit remember requests without asking again, unless sensitive.
+- **Always capture `--cwd` and `--project`** when adding a memory. Derive `--project` from the basename of cwd.
+- When a pi session is active, also capture `--session-id` and `--session-name` so memories are traceable to the conversation that produced them.
 - Ask before storing sensitive domains like finance, health, legal, identity, addresses, or private documents.
 - Never store passwords, API keys, seed phrases, OTPs, or full card numbers.
 - Prefer additive schema changes only.
 - Use `schema-change` for schema updates.
 
+## Outdated memories
+
+Memories are point-in-time snapshots. A memory recorded three months ago may no longer reflect reality — the codebase changed, the decision was reversed, or the preference shifted.
+
+When recalling memories:
+- **Check the timestamp** — `created_at` tells you when the memory was recorded. Older memories carry less weight.
+- **Verify against code** — use `read`, `bash`, `rg` to confirm a memory still reflects the current codebase before acting on it.
+- **Treat old memories as hints, not facts** — if a memory from March says "Auth uses JWT", but the current code has session cookies, the code wins.
+- **Correct stale memories** — when you detect a memory is outdated, use `add-memory` with updated info (the new memory naturally supersedes the old one in search recency). Optionally archive the old entry if needed.
+- **Surfacing staleness** — when presenting a recalled memory to the user, note its age: "I recall from [March 2026] that we decided X — let me verify that's still the case."
+
 ## When to use
+- Remember project-specific facts, decisions, conventions, or patterns
 - Save user preferences or personal facts
-- Add reminders or recurring automations
-- Update corrected memories
-- Check existing stored context before answering
+- Check existing stored context before answering a question about the current project

.pi/skills/memory/data/schema.sql

@@ -0,0 +1,85 @@
+PRAGMA foreign_keys = ON;
+
+CREATE TABLE IF NOT EXISTS memories (
+  id INTEGER PRIMARY KEY,
+  content TEXT NOT NULL,
+  category TEXT NOT NULL DEFAULT 'general',
+  context TEXT,
+  importance INTEGER NOT NULL DEFAULT 3 CHECK (importance BETWEEN 1 AND 5),
+  sensitivity TEXT NOT NULL DEFAULT 'normal' CHECK (sensitivity IN ('low','normal','high','secret')),
+  source TEXT,
+  confidence REAL NOT NULL DEFAULT 1.0 CHECK (confidence >= 0 AND confidence <= 1),
+  created_at TEXT NOT NULL DEFAULT CURRENT_TIMESTAMP,
+  updated_at TEXT NOT NULL DEFAULT CURRENT_TIMESTAMP,
+  archived INTEGER NOT NULL DEFAULT 0 CHECK (archived IN (0,1))
+);
+
+CREATE TABLE IF NOT EXISTS memory_events (
+  id INTEGER PRIMARY KEY,
+  event_type TEXT NOT NULL,
+  table_name TEXT,
+  row_id INTEGER,
+  summary TEXT,
+  raw_input TEXT,
+  created_at TEXT NOT NULL DEFAULT CURRENT_TIMESTAMP
+);
+
+CREATE TABLE IF NOT EXISTS schema_changes (
+  id INTEGER PRIMARY KEY,
+  change_type TEXT NOT NULL CHECK (change_type IN ('create_table','create_index','alter_table_add_column','other_additive')),
+  object_name TEXT NOT NULL,
+  reason TEXT NOT NULL,
+  sql TEXT NOT NULL,
+  applied_by TEXT NOT NULL DEFAULT 'agent',
+  created_at TEXT NOT NULL DEFAULT CURRENT_TIMESTAMP
+);
+
+CREATE INDEX IF NOT EXISTS idx_memories_category ON memories(category);
+CREATE INDEX IF NOT EXISTS idx_memories_context ON memories(context);
+CREATE INDEX IF NOT EXISTS idx_memories_sensitivity ON memories(sensitivity);
+CREATE INDEX IF NOT EXISTS idx_memories_importance ON memories(importance);
+CREATE INDEX IF NOT EXISTS idx_memories_archived_updated ON memories(archived, updated_at);
+CREATE INDEX IF NOT EXISTS idx_memory_events_table_row ON memory_events(table_name, row_id);
+CREATE INDEX IF NOT EXISTS idx_memory_events_created_at ON memory_events(created_at);
+CREATE INDEX IF NOT EXISTS idx_schema_changes_object ON schema_changes(object_name);
+CREATE INDEX IF NOT EXISTS idx_schema_changes_created_at ON schema_changes(created_at);
+
+CREATE VIRTUAL TABLE IF NOT EXISTS memories_fts USING fts5(
+  content,
+  category,
+  context,
+  source,
+  content='memories',
+  content_rowid='id',
+  tokenize='unicode61 remove_diacritics 2'
+);
+
+CREATE TRIGGER IF NOT EXISTS memories_fts_ai
+AFTER INSERT ON memories BEGIN
+  INSERT INTO memories_fts(rowid, content, category, context, source)
+  VALUES (new.id, new.content, new.category, new.context, new.source);
+END;
+
+CREATE TRIGGER IF NOT EXISTS memories_fts_ad
+AFTER DELETE ON memories BEGIN
+  INSERT INTO memories_fts(memories_fts, rowid, content, category, context, source)
+  VALUES('delete', old.id, old.content, old.category, old.context, old.source);
+END;
+
+CREATE TRIGGER IF NOT EXISTS memories_fts_au
+AFTER UPDATE ON memories BEGIN
+  INSERT INTO memories_fts(memories_fts, rowid, content, category, context, source)
+  VALUES('delete', old.id, old.content, old.category, old.context, old.source);
+  INSERT INTO memories_fts(rowid, content, category, context, source)
+  VALUES (new.id, new.content, new.category, new.context, new.source);
+END;
+
+CREATE TRIGGER IF NOT EXISTS memories_updated_at
+AFTER UPDATE ON memories
+FOR EACH ROW
+WHEN NEW.updated_at = OLD.updated_at
+BEGIN
+  UPDATE memories SET updated_at = CURRENT_TIMESTAMP WHERE id = OLD.id;
+END;
+
+