memor-cli 0.6.0__tar.gz → 0.7.0__tar.gz

{memor_cli-0.6.0/memor_cli.egg-info → memor_cli-0.7.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: memor-cli
-Version: 0.6.0
+Version: 0.7.0
 Summary: Measured memory for coding agents. Fire and forget — no API keys needed.
 Author-email: Nimit Bhandari <nimitbhandari17@gmail.com>
 License-Expression: MIT
@@ -46,13 +46,13 @@ Dynamic: license-file
 ```
 
 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
-[![Tests](https://img.shields.io/badge/tests-256%20passing-brightgreen.svg)]()
+[![Tests](https://img.shields.io/badge/tests-285%20passing-brightgreen.svg)]()
 [![Python](https://img.shields.io/badge/python-3.11%2B-blue.svg)]()
 [![PyPI](https://img.shields.io/pypi/v/memor-cli.svg)](https://pypi.org/project/memor-cli/)
 
-**Automatic background memory for Claude Code.** Fire and forget — no API keys needed.
+**Automatic background memory for Claude Code, Codex, and Copilot.** Fire and forget — no API keys needed.
 
-Memor watches your coding sessions, extracts decisions and patterns, and recalls relevant context on every prompt. Zero configuration. One install. Your agent remembers everything.
+Memor watches your coding sessions, extracts decisions and patterns, and recalls relevant context on every prompt. Works with Claude Code, OpenAI Codex CLI, and GitHub Copilot CLI. Zero configuration. One install. Your agent remembers everything.
 
 ---
 
@@ -62,8 +62,9 @@ Memor watches your coding sessions, extracts decisions and patterns, and recalls
 # Install globally (recommended)
 pipx install memor-cli
 
-# Install the Claude Code hook + download embedding model (~60MB)
-memor install-hook
+# Install the hook + download embedding model (~60MB)
+memor install-hook                  # interactive — pick Claude Code, Codex, or Copilot
+memor install-hook --agent claude   # or pass directly
 
 # Start as a background service (macOS/Linux)
 memor service install
@@ -72,7 +73,7 @@ memor service install
 memor daemon
 ```
 
-That's it. Every Claude Code prompt now gets automatic context recall. Open the dashboard to see it working:
+That's it. Every prompt now gets automatic context recall. Open the dashboard to see it working:
 
 ```bash
 memor dashboard
@@ -86,10 +87,10 @@ memor dashboard
 ## How It Works
 
 ```
-  You type a prompt in Claude Code
+  You type a prompt (Claude Code / Codex / Copilot)
       |
       v
-  Hook fires (UserPromptSubmit)
+  Hook fires — auto-detects which agent
       |
       v
   Embed query locally (model2vec, ~2ms)
@@ -107,14 +108,26 @@ memor dashboard
   Inject relevant context into prompt
       |
       v
-  Claude sees your past decisions, bugfixes,
+  Your agent sees past decisions, bugfixes,
   architecture choices — without you re-explaining
 ```
 
+### Supported Agents
+
+| Agent | Hook protocol | Config location | Install |
+|---|---|---|---|
+| **Claude Code** | `UserPromptSubmit` + `additionalContext` | `~/.claude/settings.json` | `memor install-hook --agent claude` |
+| **Codex CLI** | `UserPromptSubmit` + `additionalContext` | `~/.codex/hooks/hooks.json` | `memor install-hook --agent codex` |
+| **Copilot CLI** | `userPromptSubmitted` + `additionalContext` | `~/.copilot/hooks/memor.json` | `memor install-hook --agent copilot` |
+
+A single `memor-hook` binary auto-detects which agent is calling it — no separate entry points needed. The dashboard tracks recalls per agent so you can see usage across all your environments.
+
+> **Note:** Cloud-hosted agents (Codex cloud, Copilot cloud agent) run in remote sandboxes and cannot reach local hooks. MCP server support for sandboxed agents is planned ([#26](https://github.com/bnimit/memor-ai/issues/26)).
+
 **Two background processes:**
 
-1. **Daemon** — polls `~/.claude/projects/` for transcripts, embeds chunks, runs distillation, analyzes feedback (positive and negative), promotes cross-project patterns to global scope, compacts duplicates, tracks session-level token usage. All local.
-2. **Hook** — fires on every prompt, recalls relevant memories, injects them as context. Sub-15ms.
+1. **Daemon** — polls `~/.claude/projects/` for transcripts, embeds chunks, runs distillation, analyzes feedback (positive and negative), promotes cross-project patterns to global scope, compacts duplicates, auto-compacts the vector index when bloated, tracks session-level token usage. All local.
+2. **Hook** — fires on every prompt, recalls relevant memories, injects them as context. Sub-15ms. Works across Claude Code, Codex, and Copilot.
 
 **No API keys required.** Embeddings run locally via [model2vec](https://github.com/MinishLab/model2vec) (potion-base-8M, 256-dim). Vectors stored in [sqlite-vec](https://github.com/asg017/sqlite-vec). Everything runs on your machine.
 
@@ -188,10 +201,11 @@ memor dashboard
 
 Dark fintech-inspired UI showing:
 - **Hero metrics** — total memories, recall count, avg latency, coverage — with sparkline bars
+- **Agent breakdown** — per-agent recall stats (Claude, Codex, Copilot) with hit rates
 - **Daily recall activity** — stacked bar chart of hits vs misses over time
 - **Session efficiency** — real token savings measured from API usage data (avg tokens/turn with vs without recall)
 - **Per-project breakdown** — artifact counts, token totals, last activity
-- **Recent recalls** — every hook event with scores, latency, and status
+- **Recent recalls** — every hook event with agent badge, scores, latency, and status
 
 ---
 
@@ -199,7 +213,8 @@ Dark fintech-inspired UI showing:
 
 ```
 memor help                           Print the full manual
-memor install-hook                   Install Claude Code hook + download model
+memor install-hook                   Install hook + download model (interactive agent picker)
+  --agent claude|codex|copilot       Choose agent directly
 memor daemon                         Auto-ingest + distill (background watcher)
 memor dashboard                      Web dashboard on localhost:8420
 memor version                        Print installed version
@@ -211,6 +226,7 @@ memor query <text>                   Search memories from the CLI
 memor reingest                       Wipe DB and re-ingest everything
 memor reingest --project <name>      Re-ingest only one project
 memor forget-stale                   Deactivate memories unused for 30+ days
+memor compact                        Rebuild vector index, reclaim wasted space
 memor scan                           Audit DB for leaked secrets
 memor scan --purge                   Redact secrets in place
 memor setup-model                    Download/retry the embedding model
@@ -264,7 +280,8 @@ memor/
     +-- judge.py           LLM-as-judge evaluation
     +-- embed_benchmark.py Embedding model comparison
 
-memor/hook_cli.py          Claude Code hook entry point (thin client)
+memor/hook_cli.py          Hook entry point — auto-detects Claude/Codex/Copilot
+memor/hook_server.py       Hook server with agent detection + response formatting
 skill/recall.py            Standalone recall script
 ```
 
@@ -311,7 +328,7 @@ cd memor-ai
 python3 -m venv .venv && source .venv/bin/activate
 pip install -e ".[dev]"
 
-pytest  # 251 tests
+pytest  # 285 tests
 ```
 
 ---

{memor_cli-0.6.0 → memor_cli-0.7.0}/README.md

@@ -9,13 +9,13 @@
 ```
 
 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
-[![Tests](https://img.shields.io/badge/tests-256%20passing-brightgreen.svg)]()
+[![Tests](https://img.shields.io/badge/tests-285%20passing-brightgreen.svg)]()
 [![Python](https://img.shields.io/badge/python-3.11%2B-blue.svg)]()
 [![PyPI](https://img.shields.io/pypi/v/memor-cli.svg)](https://pypi.org/project/memor-cli/)
 
-**Automatic background memory for Claude Code.** Fire and forget — no API keys needed.
+**Automatic background memory for Claude Code, Codex, and Copilot.** Fire and forget — no API keys needed.
 
-Memor watches your coding sessions, extracts decisions and patterns, and recalls relevant context on every prompt. Zero configuration. One install. Your agent remembers everything.
+Memor watches your coding sessions, extracts decisions and patterns, and recalls relevant context on every prompt. Works with Claude Code, OpenAI Codex CLI, and GitHub Copilot CLI. Zero configuration. One install. Your agent remembers everything.
 
 ---
 
@@ -25,8 +25,9 @@ Memor watches your coding sessions, extracts decisions and patterns, and recalls
 # Install globally (recommended)
 pipx install memor-cli
 
-# Install the Claude Code hook + download embedding model (~60MB)
-memor install-hook
+# Install the hook + download embedding model (~60MB)
+memor install-hook                  # interactive — pick Claude Code, Codex, or Copilot
+memor install-hook --agent claude   # or pass directly
 
 # Start as a background service (macOS/Linux)
 memor service install
@@ -35,7 +36,7 @@ memor service install
 memor daemon
 ```
 
-That's it. Every Claude Code prompt now gets automatic context recall. Open the dashboard to see it working:
+That's it. Every prompt now gets automatic context recall. Open the dashboard to see it working:
 
 ```bash
 memor dashboard
@@ -49,10 +50,10 @@ memor dashboard
 ## How It Works
 
 ```
-  You type a prompt in Claude Code
+  You type a prompt (Claude Code / Codex / Copilot)
       |
       v
-  Hook fires (UserPromptSubmit)
+  Hook fires — auto-detects which agent
       |
       v
   Embed query locally (model2vec, ~2ms)
@@ -70,14 +71,26 @@ memor dashboard
   Inject relevant context into prompt
       |
       v
-  Claude sees your past decisions, bugfixes,
+  Your agent sees past decisions, bugfixes,
   architecture choices — without you re-explaining
 ```
 
+### Supported Agents
+
+| Agent | Hook protocol | Config location | Install |
+|---|---|---|---|
+| **Claude Code** | `UserPromptSubmit` + `additionalContext` | `~/.claude/settings.json` | `memor install-hook --agent claude` |
+| **Codex CLI** | `UserPromptSubmit` + `additionalContext` | `~/.codex/hooks/hooks.json` | `memor install-hook --agent codex` |
+| **Copilot CLI** | `userPromptSubmitted` + `additionalContext` | `~/.copilot/hooks/memor.json` | `memor install-hook --agent copilot` |
+
+A single `memor-hook` binary auto-detects which agent is calling it — no separate entry points needed. The dashboard tracks recalls per agent so you can see usage across all your environments.
+
+> **Note:** Cloud-hosted agents (Codex cloud, Copilot cloud agent) run in remote sandboxes and cannot reach local hooks. MCP server support for sandboxed agents is planned ([#26](https://github.com/bnimit/memor-ai/issues/26)).
+
 **Two background processes:**
 
-1. **Daemon** — polls `~/.claude/projects/` for transcripts, embeds chunks, runs distillation, analyzes feedback (positive and negative), promotes cross-project patterns to global scope, compacts duplicates, tracks session-level token usage. All local.
-2. **Hook** — fires on every prompt, recalls relevant memories, injects them as context. Sub-15ms.
+1. **Daemon** — polls `~/.claude/projects/` for transcripts, embeds chunks, runs distillation, analyzes feedback (positive and negative), promotes cross-project patterns to global scope, compacts duplicates, auto-compacts the vector index when bloated, tracks session-level token usage. All local.
+2. **Hook** — fires on every prompt, recalls relevant memories, injects them as context. Sub-15ms. Works across Claude Code, Codex, and Copilot.
 
 **No API keys required.** Embeddings run locally via [model2vec](https://github.com/MinishLab/model2vec) (potion-base-8M, 256-dim). Vectors stored in [sqlite-vec](https://github.com/asg017/sqlite-vec). Everything runs on your machine.
 
@@ -151,10 +164,11 @@ memor dashboard
 
 Dark fintech-inspired UI showing:
 - **Hero metrics** — total memories, recall count, avg latency, coverage — with sparkline bars
+- **Agent breakdown** — per-agent recall stats (Claude, Codex, Copilot) with hit rates
 - **Daily recall activity** — stacked bar chart of hits vs misses over time
 - **Session efficiency** — real token savings measured from API usage data (avg tokens/turn with vs without recall)
 - **Per-project breakdown** — artifact counts, token totals, last activity
-- **Recent recalls** — every hook event with scores, latency, and status
+- **Recent recalls** — every hook event with agent badge, scores, latency, and status
 
 ---
 
@@ -162,7 +176,8 @@ Dark fintech-inspired UI showing:
 
 ```
 memor help                           Print the full manual
-memor install-hook                   Install Claude Code hook + download model
+memor install-hook                   Install hook + download model (interactive agent picker)
+  --agent claude|codex|copilot       Choose agent directly
 memor daemon                         Auto-ingest + distill (background watcher)
 memor dashboard                      Web dashboard on localhost:8420
 memor version                        Print installed version
@@ -174,6 +189,7 @@ memor query <text>                   Search memories from the CLI
 memor reingest                       Wipe DB and re-ingest everything
 memor reingest --project <name>      Re-ingest only one project
 memor forget-stale                   Deactivate memories unused for 30+ days
+memor compact                        Rebuild vector index, reclaim wasted space
 memor scan                           Audit DB for leaked secrets
 memor scan --purge                   Redact secrets in place
 memor setup-model                    Download/retry the embedding model
@@ -227,7 +243,8 @@ memor/
     +-- judge.py           LLM-as-judge evaluation
     +-- embed_benchmark.py Embedding model comparison
 
-memor/hook_cli.py          Claude Code hook entry point (thin client)
+memor/hook_cli.py          Hook entry point — auto-detects Claude/Codex/Copilot
+memor/hook_server.py       Hook server with agent detection + response formatting
 skill/recall.py            Standalone recall script
 ```
 
@@ -274,7 +291,7 @@ cd memor-ai
 python3 -m venv .venv && source .venv/bin/activate
 pip install -e ".[dev]"
 
-pytest  # 251 tests
+pytest  # 285 tests
 ```
 
 ---

memor_cli-0.7.0/memor/__init__.py

@@ -0,0 +1 @@
+__version__ = "0.7.0"

{memor_cli-0.6.0 → memor_cli-0.7.0}/memor/cli.py

@@ -44,9 +44,10 @@ HELP_TEXT = """\
 memor — measured memory for coding agents
 
 GETTING STARTED
-  memor install-hook        Install the Claude Code recall hook + download model
-  memor service install     Start the daemon as a background service
-  memor dashboard           Open the web dashboard at localhost:8420
+  memor install-hook            Install the recall hook + download model
+    --agent claude|codex|copilot  Choose your agent (default: claude)
+  memor service install         Start the daemon as a background service
+  memor dashboard               Open the web dashboard at localhost:8420
 
 SERVICE MANAGEMENT
   memor service install     Install and start as a background service (survives reboots)
@@ -72,6 +73,7 @@ MAINTENANCE
   memor reingest --project <name>  Re-ingest only one project
   memor distill --project <name>   Run distillation manually
   memor forget-stale               Deactivate memories not recalled in 30 days
+  memor compact                    Rebuild vector index, reclaim space
   memor scan                       Audit DB for leaked secrets
   memor scan --purge               Redact secrets in place
   memor setup-model                Download/retry the embedding model (~60MB)
@@ -88,7 +90,9 @@ CONFIGURATION
   No configuration needed — just install, hook, and run the daemon.
 
 EXAMPLES
-  memor install-hook && memor service install  # one-time setup
+  memor install-hook && memor service install          # Claude Code (default)
+  memor install-hook --agent codex && memor service install  # Codex CLI
+  memor install-hook --agent copilot && memor service install # Copilot CLI
   memor query "auth flow" --project my-app
   memor reingest --project my-app -y           # refresh one project
   memor dashboard                              # see stats at localhost:8420
@@ -391,6 +395,36 @@ def forget_stale(days: int = typer.Option(30, help="Deactivate memories not reca
     typer.echo(f"Deactivated {count} stale memories.")
 
 
+@app.command("compact")
+def compact(db: str = typer.Option(str(Path.home() / ".memor" / "memor.db")),
+            fake: bool = False,
+            confirm: bool = typer.Option(False, "--yes", "-y", help="Skip confirmation")):
+    """Rebuild the vector index to reclaim wasted space."""
+    db_path = _db_path(db)
+    if not Path(db_path).exists():
+        typer.echo("No database found.")
+        raise typer.Exit(1)
+    dim = _get_dim(db_path)
+    embedder = _embedder(fake)
+    if embedder.dim != dim:
+        typer.echo(f"Warning: embedder dim={embedder.dim} != DB dim={dim}. Using DB dim.")
+        from memor.embed.fake import FakeEmbedder
+        embedder = FakeEmbedder(dim=dim)
+    store = SqliteStore(db_path, dim=dim)
+    chunk_count = store.db.execute("SELECT COUNT(*) as c FROM vec_artifacts_chunks").fetchone()["c"]
+    active_count = store.db.execute("SELECT COUNT(*) as c FROM artifacts WHERE active=1").fetchone()["c"]
+    db_size_mb = round(Path(db_path).stat().st_size / 1_048_576, 1)
+    typer.echo(f"Current: {chunk_count} chunks, {active_count} active vectors, {db_size_mb} MB")
+    if not confirm:
+        typer.confirm("Rebuild vector index?", abort=True)
+    typer.echo("Compacting...")
+    result = store.rebuild_vec_index(embedder)
+    db_size_after = round(Path(db_path).stat().st_size / 1_048_576, 1)
+    typer.echo(f"Done: {result['vectors_reindexed']} vectors reindexed in {result['duration_ms']}ms")
+    typer.echo(f"Chunks: {result['before_chunks']} -> {result['after_chunks']} (chunk_size={result['chunk_size']})")
+    typer.echo(f"DB size: {db_size_mb} MB -> {db_size_after} MB")
+
+
 @app.command("scan")
 def scan(db: str = typer.Option(str(Path.home() / ".memor" / "memor.db")),
          purge: bool = typer.Option(False, "--purge", help="Redact secrets in place"),
@@ -456,20 +490,108 @@ def _install_hook_logic(settings_path: Path, hook_command: str) -> None:
     settings_path.write_text(json.dumps(data, indent=2))
 
 
+def _install_hook_logic_codex(hooks_path: Path, hook_command: str) -> None:
+    if hooks_path.exists():
+        data = json.loads(hooks_path.read_text())
+    else:
+        data = {}
+    hooks = data.setdefault("hooks", {})
+    prompt_hooks = hooks.setdefault("UserPromptSubmit", [])
+    entry = {"command": hook_command, "timeout": 5}
+    existing_idx = None
+    for i, h in enumerate(prompt_hooks):
+        if "memor-hook" in h.get("command", ""):
+            existing_idx = i
+            break
+    if existing_idx is not None:
+        prompt_hooks[existing_idx] = entry
+    else:
+        prompt_hooks.append(entry)
+    hooks_path.parent.mkdir(parents=True, exist_ok=True)
+    hooks_path.write_text(json.dumps(data, indent=2))
+
+
+def _install_hook_logic_copilot(hooks_path: Path, hook_command: str) -> None:
+    if hooks_path.exists():
+        data = json.loads(hooks_path.read_text())
+    else:
+        data = {"version": 1, "hooks": {}}
+    data.setdefault("version", 1)
+    hooks = data.setdefault("hooks", {})
+    prompt_hooks = hooks.setdefault("userPromptSubmitted", [])
+    entry = {"type": "command", "bash": hook_command, "timeoutSec": 5}
+    existing_idx = None
+    for i, h in enumerate(prompt_hooks):
+        if "memor-hook" in h.get("bash", ""):
+            existing_idx = i
+            break
+    if existing_idx is not None:
+        prompt_hooks[existing_idx] = entry
+    else:
+        prompt_hooks.append(entry)
+    hooks_path.parent.mkdir(parents=True, exist_ok=True)
+    hooks_path.write_text(json.dumps(data, indent=2))
+
+
+AGENT_CHOICES = [
+    ("claude", "~/.claude/settings.json", "Claude Code"),
+    ("codex", "~/.codex/hooks/hooks.json", "Codex CLI"),
+    ("copilot", "~/.copilot/hooks/memor.json", "Copilot CLI"),
+]
+
+
+def _prompt_agent() -> str:
+    typer.echo("Which agent do you want to install the recall hook for?\n")
+    for i, (key, _, label) in enumerate(AGENT_CHOICES, 1):
+        typer.echo(f"  {i}. {label}")
+    typer.echo()
+    while True:
+        choice = typer.prompt("Choose (1-3)", default="1")
+        try:
+            idx = int(choice) - 1
+            if 0 <= idx < len(AGENT_CHOICES):
+                return AGENT_CHOICES[idx][0]
+        except ValueError:
+            pass
+        typer.echo("Invalid choice. Enter 1, 2, or 3.")
+
+
 @app.command("install-hook")
-def install_hook():
-    """Install the Claude Code recall hook into ~/.claude/settings.json."""
+def install_hook(agent: str | None = typer.Option(None, help="Agent: claude, codex, copilot (interactive if omitted)")):
+    """Install the recall hook for Claude Code, Codex, or Copilot."""
     import shutil
+    agent_keys = {key for key, _, _ in AGENT_CHOICES}
+
+    if agent is None:
+        agent = _prompt_agent()
+    else:
+        agent = agent.lower()
+
+    if agent not in agent_keys:
+        typer.echo(f"Unknown agent '{agent}'. Choose: claude, codex, copilot", err=True)
+        raise typer.Exit(1)
+
     hook_bin = shutil.which("memor-hook")
     if not hook_bin:
         typer.echo("Error: 'memor-hook' not found on PATH.", err=True)
         typer.echo("  If you installed with pipx: pipx install memor-cli", err=True)
         typer.echo("  If developing locally: pip install -e .", err=True)
         raise typer.Exit(1)
-    settings_path = Path.home() / ".claude" / "settings.json"
-    _install_hook_logic(settings_path, hook_bin)
-    typer.echo(f"Hook installed: {hook_bin}")
-    typer.echo(f"Settings updated: {settings_path}")
+
+    config_path_str, agent_name = next(
+        (p, n) for k, p, n in AGENT_CHOICES if k == agent
+    )
+    config_path = Path(config_path_str).expanduser()
+
+    if agent == "claude":
+        _install_hook_logic(config_path, hook_bin)
+    elif agent == "codex":
+        _install_hook_logic_codex(config_path, hook_bin)
+    elif agent == "copilot":
+        _install_hook_logic_copilot(config_path, hook_bin)
+
+    typer.echo(f"\nHook installed for {agent_name}: {hook_bin}")
+    typer.echo(f"Config updated: {config_path}")
     typer.echo()
     typer.echo("Pre-downloading embedding model...")
     try:

{memor_cli-0.6.0 → memor_cli-0.7.0}/memor/daemon.py

@@ -186,6 +186,29 @@ def compact_memories(store: SqliteStore, embedder) -> int:
     return deactivated
 
 
+def should_compact(store: SqliteStore) -> bool:
+    try:
+        chunk_count = store.db.execute(
+            "SELECT COUNT(*) as c FROM vec_artifacts_chunks").fetchone()["c"]
+        active_count = store.db.execute(
+            "SELECT COUNT(*) as c FROM artifacts WHERE active=1").fetchone()["c"]
+    except Exception as e:
+        print(f"  should_compact error: {e}")
+        return False
+    if chunk_count == 0 or active_count == 0:
+        return False
+    from memor.store.sqlite_store import _choose_chunk_size
+    chunk_size = _choose_chunk_size(active_count)
+    ideal_chunks = max(1, active_count // chunk_size + 1)
+    return chunk_count > ideal_chunks * 2
+
+
+def auto_compact(store: SqliteStore, embedder) -> dict | None:
+    if not should_compact(store):
+        return None
+    return store.rebuild_vec_index(embedder, vacuum=False)
+
+
 def run_poll_cycle(
     state: dict[str, float],
     store: SqliteStore,
@@ -286,6 +309,16 @@ def run_poll_cycle(
         except Exception:
             pass
 
+    # Auto-compact vec index if bloated
+    if new_ingested:
+        try:
+            result = auto_compact(store, embedder)
+            if result:
+                print(f"  vec compact: {result['before_chunks']} -> {result['after_chunks']} chunks "
+                      f"({result['vectors_reindexed']} vectors, {result['duration_ms']}ms)")
+        except Exception:
+            pass
+
     return state, distilled
 
 

{memor_cli-0.6.0 → memor_cli-0.7.0}/memor/dashboard/server.py

@@ -14,9 +14,13 @@ def create_app(db_path: str | None = None) -> FastAPI:
 
     app = FastAPI(title="Memor Dashboard")
     _db_path = db_path
+    _cached_store: SqliteStore | None = None
 
     def _store() -> SqliteStore:
-        return SqliteStore(_db_path, dim=_get_dim(_db_path))
+        nonlocal _cached_store
+        if _cached_store is None:
+            _cached_store = SqliteStore(_db_path, dim=_get_dim(_db_path))
+        return _cached_store
 
     @app.get("/", response_class=HTMLResponse)
     def index():
@@ -66,11 +70,16 @@ def create_app(db_path: str | None = None) -> FastAPI:
             ORDER BY artifacts DESC
         """).fetchall()
 
+        active_rows = [
+            r for r in artifact_rows
+            if (r["memories"] or 0) > 0 or r["project"] in recall_projects
+        ]
+
         if not recall_stats:
-            return [dict(r) for r in artifact_rows]
+            return [dict(r) for r in active_rows]
 
         result = list(recall_stats)
-        for r in artifact_rows:
+        for r in active_rows:
             if r["project"] not in recall_projects:
                 result.append(dict(r))
         return result
@@ -168,6 +177,11 @@ def create_app(db_path: str | None = None) -> FastAPI:
             return {"status": "no_runs"}
         return result
 
+    @app.get("/api/agent-breakdown")
+    def agent_breakdown():
+        store = _store()
+        return store.get_agent_breakdown()
+
     @app.get("/api/version")
     def version():
         from memor import __version__

{memor_cli-0.6.0 → memor_cli-0.7.0}/memor/dashboard/static/index.html

@@ -304,6 +304,12 @@
   .badge-global::before { background: #a090ff; }
   .badge-negative { background: rgba(232,84,84,0.12); color: var(--error); }
   .badge-negative::before { background: var(--error); }
+  .badge-claude { background: var(--accent-dim); color: var(--accent); }
+  .badge-claude::before { background: var(--accent); }
+  .badge-codex { background: rgba(61,214,140,0.12); color: #3dd68c; }
+  .badge-codex::before { background: #3dd68c; }
+  .badge-copilot { background: rgba(100,160,255,0.12); color: #64a0ff; }
+  .badge-copilot::before { background: #64a0ff; }
 
   .quality-bar { display: flex; height: 6px; border-radius: 3px; overflow: hidden; background: var(--surface3); width: 100%; min-width: 60px; }
   .quality-fill { height: 100%; border-radius: 3px; transition: width 0.3s; }
@@ -398,6 +404,12 @@
     </div>
   </section>
 
+  <!-- ── Agent breakdown ─────────────────────────────────── -->
+  <section id="agent-section" style="display:none;">
+    <div class="section-label"><span class="dot"></span> Agent Breakdown</div>
+    <div class="metric-grid" id="agent-grid" style="grid-template-columns:repeat(auto-fit,minmax(180px,1fr));"></div>
+  </section>
+
   <!-- ── Token ROI hero ─────────────────────────────────── -->
   <section>
     <div class="chart-card" id="roi-card" style="display:none;">
@@ -538,6 +550,7 @@
         <thead>
           <tr>
             <th>Time</th>
+            <th>Agent</th>
             <th>Project</th>
             <th>Query</th>
             <th>Hits</th>
@@ -595,6 +608,16 @@
     return '<span class="badge ' + m[0] + '">' + m[1] + '</span>';
   }
 
+  function agentBadge(agent) {
+    var map = {
+      claude:  ['badge-claude',  'Claude'],
+      codex:   ['badge-codex',   'Codex'],
+      copilot: ['badge-copilot', 'Copilot'],
+    };
+    var m = map[agent || 'claude'] || ['badge-claude', agent || 'claude'];
+    return '<span class="badge ' + m[0] + '">' + m[1] + '</span>';
+  }
+
   function el(tag, attrs, inner) {
     var e = document.createElement(tag);
     for (var k in (attrs || {})) e.setAttribute(k, attrs[k]);
@@ -860,13 +883,14 @@
     var tbody = document.getElementById('recalls-body');
     tbody.innerHTML = '';
     if (!data.length) {
-      tbody.innerHTML = '<tr><td colspan="8" class="empty-state">No recall events yet.</td></tr>';
+      tbody.innerHTML = '<tr><td colspan="9" class="empty-state">No recall events yet.</td></tr>';
       return;
     }
     data.forEach(function(row) {
       var tr = el('tr', {});
       tr.innerHTML =
         '<td class="mono" style="white-space:nowrap;color:var(--text-muted)">' + relativeTime(row.timestamp) + '</td>' +
+        '<td>' + agentBadge(row.agent) + '</td>' +
         '<td class="truncate mono" style="max-width:120px" title="' + esc(row.project) + '">' + esc(row.project) + '</td>' +
         '<td class="truncate" style="max-width:220px" title="' + esc(row.query_preview) + '">' + esc(row.query_preview) + '</td>' +
         '<td>' + (row.hits_count||0) + '</td>' +
@@ -1002,6 +1026,29 @@
     });
   }
 
+  /* ── Agent breakdown ────────────────────────────────────── */
+  function renderAgentBreakdown(data) {
+    var section = document.getElementById('agent-section');
+    var grid = document.getElementById('agent-grid');
+    if (!data || data.length <= 1) { section.style.display = 'none'; return; }
+    section.style.display = '';
+    grid.innerHTML = '';
+    var colors = { claude: 'var(--accent)', codex: '#3dd68c', copilot: '#64a0ff' };
+    data.forEach(function(row) {
+      var agent = row.agent || 'claude';
+      var color = colors[agent] || 'var(--text-muted)';
+      var hitRate = row.recalls > 0 ? ((row.hits / row.recalls) * 100) : 0;
+      var card = el('div', {class: 'metric-card'});
+      card.innerHTML =
+        '<div class="metric-info">' +
+          '<div class="metric-label" style="color:' + color + ';">' + esc(agent.charAt(0).toUpperCase() + agent.slice(1)) + '</div>' +
+          '<div class="metric-value">' + fmt(row.recalls) + '</div>' +
+          '<div class="metric-sub">' + fmtPct(hitRate) + ' hit rate · ' + fmtMs(row.avg_latency) + '</div>' +
+        '</div>';
+      grid.appendChild(card);
+    });
+  }
+
   /* ── Data loaders ──────────────────────────────────────── */
   async function loadSummary()    { try { renderSummary(await api('/api/summary')); } catch(e) { console.warn('summary',e); } }
   async function loadProjects()   { try { renderProjects(await api('/api/projects')); } catch(e) { console.warn('projects',e); } }
@@ -1012,6 +1059,7 @@
   async function loadROI()        { try { renderROI(await api('/api/roi')); } catch(e) { console.warn('roi',e); } }
   async function loadROITrend()   { try { renderROITrend(await api('/api/roi-trend')); } catch(e) { console.warn('roi-trend',e); } }
   async function loadQuality()    { try { renderQuality(await api('/api/quality')); } catch(e) { console.warn('quality',e); } }
+  async function loadAgentBreakdown() { try { renderAgentBreakdown(await api('/api/agent-breakdown')); } catch(e) { console.warn('agent-breakdown',e); } }
   async function loadRecalls() {
     try {
       var url = '/api/recalls?limit=50' + (projectFilter ? '&project=' + encodeURIComponent(projectFilter) : '');
@@ -1020,7 +1068,7 @@
   }
 
   async function refresh() {
-    await Promise.allSettled([loadSummary(), loadProjects(), loadEfficiency(), loadSessionEfficiency(), loadHealth(), loadTrend(), loadRecalls(), loadROI(), loadROITrend(), loadQuality()]);
+    await Promise.allSettled([loadSummary(), loadProjects(), loadEfficiency(), loadSessionEfficiency(), loadHealth(), loadTrend(), loadRecalls(), loadROI(), loadROITrend(), loadQuality(), loadAgentBreakdown()]);
     document.getElementById('last-updated').textContent = new Date().toLocaleTimeString();
 
     renderMiniBars('mb-chunks', null);