@goondocks/myco 0.2.9 → 0.2.11

package/.claude-plugin/marketplace.json

@@ -11,11 +11,19 @@
       "name": "myco",
       "source": {
         "source": "npm",
-        "package": "@goondocks/myco"
+        "package": "@goondocks/myco",
+        "version": "0.2.10"
       },
       "description": "Collective agent intelligence — captures session knowledge and serves it back via MCP",
       "license": "MIT",
-      "keywords": ["intelligence", "memory", "mcp", "sessions", "team", "knowledge"],
+      "keywords": [
+        "intelligence",
+        "memory",
+        "mcp",
+        "sessions",
+        "team",
+        "knowledge"
+      ],
       "category": "productivity"
     }
   ]

package/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "myco",
-  "version": "0.2.9",
+  "version": "0.2.11",
   "description": "Collective agent intelligence — captures session knowledge and serves it back to your team via MCP",
   "author": {
     "name": "goondocks-co",

package/commands/init.md

@@ -5,9 +5,9 @@ description: Initialize Myco in the current project — sets up vault, config, a
 
 # Initialize Myco
 
-Set up Myco for this project. Guide the user through:
+Guide the user through setup, then run the CLI to create the vault. **Do NOT create files manually — the CLI handles all vault creation, config writing, and env configuration.**
 
-## Step 0: Choose vault location
+## Step 1: Choose vault location
 
 Ask the user where they want the vault:
 
@@ -17,199 +17,50 @@ Ask the user where they want the vault:
 > 2. **Centralized** (`~/.myco/vaults/<project-name>/`) — vault stays outside the repo, good for public repos or personal use
 > 3. **Custom path** — specify your own location
 
-Pass the chosen path to the CLI via `--vault <path>`. The CLI handles all vault setup, env configuration, and agent detection.
-
-## Step 1: Create vault directory
-
-Create the vault directory (at the resolved path from Step 0) with subdirectories:
-`sessions`, `plans`, `memories`, `artifacts`, `team`, `buffer`, `logs`
-
-Also create a `_dashboard.md` file in the vault root with the following Dataview-powered content:
-
-```markdown
-# Myco Vault
-
-## Active Plans
-\`\`\`dataview
-TABLE status, tags FROM #type/plan
-WHERE status = "active" OR status = "in_progress"
-SORT created DESC
-\`\`\`
-
-## Recent Sessions
-\`\`\`dataview
-TABLE user, started, tools_used FROM #type/session
-SORT started DESC LIMIT 10
-\`\`\`
-
-## Recent Memories
-\`\`\`dataview
-TABLE observation_type AS "Type", created FROM #type/memory
-SORT created DESC LIMIT 15
-\`\`\`
-
-## Memories by Type
-\`\`\`dataview
-TABLE WITHOUT ID observation_type AS "Type", length(rows) AS "Count"
-FROM #type/memory GROUP BY observation_type
-SORT length(rows) DESC
-\`\`\`
-
-## Gotchas
-\`\`\`dataview
-LIST FROM #memory/gotcha SORT created DESC LIMIT 10
-\`\`\`
-```
-
-This dashboard requires the Dataview community plugin in Obsidian. Without it, the code blocks are visible but still readable as plain markdown.
-
 ## Step 2: Choose intelligence backend
 
-Configure LLM and embedding providers independently:
+Detect available providers by checking local endpoints:
 
-### LLM provider
+- **Ollama** — `curl -s http://localhost:11434/api/tags` — list model names
+- **LM Studio** — `curl -s http://localhost:1234/v1/models` — list model IDs
+- **Anthropic** — check if `ANTHROPIC_API_KEY` is set
 
-Ask the user to choose an LLM provider:
+Show the user what's available and recommend:
+- **LLM**: `gpt-oss` on Ollama or LM Studio (best for structured JSON output)
+- **Embeddings**: `bge-m3` on Ollama (Anthropic does not support embeddings)
 
-- **Ollama** — detect at `http://localhost:11434/api/tags`, list available models
-- **LM Studio** — detect at `http://localhost:1234/v1/models`, list available models
-- **Anthropic** — uses existing `ANTHROPIC_API_KEY`, verify it's set
-
-Recommended summarization models by hardware tier:
-
-| Tier | Models | RAM | Notes |
-|------|--------|-----|-------|
-| **High** (best quality) | `gpt-oss` (~20B), `gemma3:27b`, `qwen3.5:14b` | 16GB+ | Best observation extraction and structured JSON output |
-| **Mid** (good balance) | `qwen3.5:8b`, `gemma3:12b` | 8GB+ | Good quality, reasonable speed |
-| **Light** (resource constrained) | `gemma3:4b`, `qwen3.5:4b` | 4GB+ | Faster, may miss nuanced observations |
-
-If the user already has a model loaded, prefer using what they have — any instruction-tuned model that handles JSON output well will work. The model only needs to produce structured JSON (observation extraction) and short text (summaries, titles).
-
-For the selected provider, list available models and let the user choose. Also set:
-- `context_window` (default 8192) — only for local providers, not Anthropic
-- `max_tokens` (default 1024)
+Let the user choose their LLM provider/model and embedding provider/model.
 
 If the recommended model isn't available, offer to pull it:
-- **Ollama**: `ollama pull gpt-oss` (pulls latest tag automatically)
-- **LM Studio**: `lms get openai/gpt-oss-20b` (uses `owner/model` format)
-
-Ask the user before pulling — models can be large (hundreds of MB to several GB).
-
-### Embedding provider
-
-Ask the user to choose an embedding provider. **Anthropic is not an option here** — it doesn't support embeddings.
-
-- **Ollama** — detect at `http://localhost:11434/api/tags`, list available models, recommend `bge-m3` or `nomic-embed-text`. Ollama is the recommended provider for embeddings.
-- **LM Studio** — possible but not recommended for embeddings. LM Studio is better suited for LLM/summarization work.
-
-For the selected provider, list available models and let the user choose.
-
-If the recommended embedding model isn't installed, offer to pull it — embedding models are typically small (~300-700MB):
-- **Ollama**: `ollama pull bge-m3`
-
-## Step 3: Team / solo setup
-
-Ask whether this is a team or solo project:
-
-- **Solo** — vault stays local, not tracked by git
-- **Team** — set up git tracking for the vault directory, ask for username
-
-If `MYCO_VAULT_DIR` is set in the environment, also offer:
-- **Use MYCO_VAULT_DIR from env** — treat the env-specified vault as a shared/external vault managed outside this repo; skip git tracking
-
-## Step 4: Write `myco.yaml`
-
-Write a `version: 2` config file with chosen settings. **All configurable values must be explicit** — no hidden schema defaults. Example output:
-
-```yaml
-version: 2
-
-intelligence:
-  llm:
-    provider: ollama
-    model: gpt-oss
-    base_url: http://localhost:11434
-    context_window: 8192
-    max_tokens: 1024
-  embedding:
-    provider: ollama
-    model: bge-m3
-    base_url: http://localhost:11434
-
-daemon:
-  log_level: info
-  grace_period: 30
-  max_log_size: 5242880
-
-capture:
-  transcript_paths: []
-  artifact_watch:
-    - .claude/plans/
-    - .cursor/plans/
-  artifact_extensions:
-    - .md
-  buffer_max_events: 500
-
-context:
-  max_tokens: 1200
-  layers:
-    plans: 200
-    sessions: 500
-    memories: 300
-    team: 200
-
-team:
-  enabled: false
-  user: ""
-  sync: git
+- **Ollama**: `ollama pull <model>`
+- **LM Studio**: `lms get <owner/model>`
+
+## Step 3: Run the CLI
+
+Run the init command with all gathered inputs. The CLI creates the vault, writes config, sets up the FTS index, and configures `MYCO_VAULT_DIR` if the vault is external:
+
+```bash
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js init \
+  --vault <chosen-path> \
+  --llm-provider <provider> \
+  --llm-model <model> \
+  --llm-url <base-url> \
+  --embedding-provider <provider> \
+  --embedding-model <model> \
+  --embedding-url <base-url>
 ```
 
-Substitute the user's chosen providers, models, and base URLs. Set `team.enabled`, `team.user`, and `team.sync` based on Step 3.
+## Step 4: Verify
 
-## Step 5: Write vault `.gitignore`
+After the CLI completes, confirm providers are reachable:
 
-Create a `.gitignore` inside the `.myco/` vault directory to exclude runtime artifacts while committing the knowledge:
-
-```
-# Runtime — rebuilt on daemon startup
-index.db
-index.db-wal
-index.db-shm
-vectors.db
-
-# Daemon state — per-machine, ephemeral
-daemon.json
-buffer/
-logs/
-
-# Obsidian — per-user workspace config
-.obsidian/
-```
-
-Everything else is committed: `myco.yaml`, `sessions/`, `memories/`, `plans/`, `artifacts/`, `team/`, `lineage.json`, `_dashboard.md`. This is the project's institutional memory — it travels with the code.
-
-## Step 6: Vault discovery and MCP
-
-The `MYCO_VAULT_DIR` env var (if needed) was already set in Step 0. No additional configuration is required.
-
-**Cursor / VS Code** — if the user chose an external vault path, instruct them to also set `MYCO_VAULT_DIR` in their shell profile (`~/.zshrc`, `~/.bashrc`) so other agents can find it.
-
-All three agents (Claude Code, Cursor, VS Code Copilot) auto-discover the MCP server from the plugin manifest when installed via the marketplace. No manual `.mcp.json` editing is needed.
-
-## Step 7: Setup summary
-
-After setup, display a summary:
+1. Test the LLM — send a short prompt and verify a response
+2. Test embeddings — generate a test embedding and report dimensions
+3. Display a setup summary table
 
 | Setting | Value |
 |---------|-------|
-| Vault path | `<resolved path>` (`<vault path source>`) |
+| Vault path | `<resolved path>` |
 | LLM provider | `<provider>` / `<model>` |
 | Embedding provider | `<provider>` / `<model>` |
 | Context window | `<context_window>` |
-| Team mode | `<enabled/disabled>` |
-
-Then confirm everything is working:
-1. Verify the LLM provider is reachable (call `isAvailable()`)
-2. Verify the embedding provider is reachable (call `isAvailable()`)
-3. Run a test embedding to confirm dimensions
-4. Report success or issues found

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@goondocks/myco",
-  "version": "0.2.9",
+  "version": "0.2.11",
   "description": "Collective agent intelligence — Claude Code plugin",
   "type": "module",
   "main": "dist/index.js",