@goondocks/myco 0.4.1 → 0.4.2

package/skills/myco/references/cli-usage.md

@@ -0,0 +1,322 @@
+# Myco CLI Reference
+
+All CLI commands are invoked as:
+
+    node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js <command> [flags]
+
+Where `${CLAUDE_PLUGIN_ROOT}` resolves to the plugin root directory (set by the agent automatically).
+
+---
+
+## Setup Commands
+
+### `init` — Create vault structure and base config
+
+Initializes a new Myco vault. Skipped automatically if the vault is already initialized.
+
+| Flag | Type | Description |
+|------|------|-------------|
+| `--vault <path>` | string | Custom vault directory (supports `~/` expansion) |
+| `--llm-provider <name>` | string | `ollama`, `lm-studio`, or `anthropic` |
+| `--llm-model <name>` | string | Model name |
+| `--llm-url <url>` | string | Provider base URL |
+| `--embedding-provider <name>` | string | `ollama` or `lm-studio` |
+| `--embedding-model <name>` | string | Embedding model name |
+| `--embedding-url <url>` | string | Embedding provider base URL |
+| `--user <name>` | string | Username for team-enabled vault |
+| `--team` | boolean | Enable team collaboration mode |
+| `--tiers <csv>` | string | Comma-separated digest tier list (e.g., `1500,3000,5000`) |
+| `--inject-tier <number>` | number | Tier to auto-inject at session start |
+| `--context-window <number>` | number | Context window size for digest operations |
+
+**Example:**
+
+```sh
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js init \
+  --llm-provider ollama \
+  --llm-model qwen2.5-coder:14b \
+  --embedding-provider ollama \
+  --embedding-model nomic-embed-text
+```
+
+---
+
+### `setup-llm` — Change LLM and embedding provider settings
+
+Reconfigures intelligence backend without reinitializing the vault.
+
+| Flag | Type | Description |
+|------|------|-------------|
+| `--llm-provider <name>` | string | Provider name (`ollama`, `lm-studio`, `anthropic`) |
+| `--llm-model <name>` | string | Model name |
+| `--llm-url <url>` | string | Provider base URL |
+| `--llm-context-window <number>` | number | Context window in tokens |
+| `--llm-max-tokens <number>` | number | Max output tokens |
+| `--embedding-provider <name>` | string | Embedding provider name |
+| `--embedding-model <name>` | string | Embedding model name |
+| `--embedding-url <url>` | string | Embedding base URL |
+| `--show` | boolean | Display current settings and exit |
+
+Note: changing the embedding model requires running `rebuild` afterward to re-embed all vault notes with the new model.
+
+**Example:**
+
+```sh
+# Show current settings
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js setup-llm --show
+
+# Switch to Anthropic
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js setup-llm \
+  --llm-provider anthropic \
+  --llm-model claude-opus-4-5
+```
+
+---
+
+### `setup-digest` — Configure digest and capture settings
+
+Controls the continuous reasoning system: metabolism timing, tier configuration, and per-operation token budgets.
+
+| Flag | Type | Description |
+|------|------|-------------|
+| `--enabled <true\|false>` | boolean | Enable or disable digest |
+| `--tiers <csv>` | string | Comma-separated token tier list |
+| `--inject-tier <number\|null>` | number\|null | Auto-inject tier at session start (`null` to disable) |
+| `--provider <name\|null>` | string\|null | Digest-specific provider (`null` = inherit from main LLM) |
+| `--model <name\|null>` | string\|null | Digest-specific model (`null` = inherit) |
+| `--base-url <url\|null>` | string\|null | Digest provider base URL (`null` = inherit) |
+| `--context-window <number>` | number | Context window for digest |
+| `--keep-alive <duration>` | string | Keep model loaded (Ollama only, e.g., `30m`) |
+| `--gpu-kv-cache <true\|false>` | boolean | GPU KV cache offload (LM Studio only) |
+| `--active-interval <seconds>` | number | Metabolism active processing interval |
+| `--dormancy-threshold <seconds>` | number | Time before entering dormancy |
+| `--max-notes <number>` | number | Max substrate notes per digest cycle |
+| `--extraction-tokens <number>` | number | Max tokens for spore extraction |
+| `--summary-tokens <number>` | number | Max tokens for session summaries |
+| `--title-tokens <number>` | number | Max tokens for session titles |
+| `--classification-tokens <number>` | number | Max tokens for artifact classification |
+| `--show` | boolean | Display current settings and exit |
+
+**Example:**
+
+```sh
+# Show current digest config
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js setup-digest --show
+
+# Use a faster model just for digest, with shorter active interval
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js setup-digest \
+  --model qwen2.5:7b \
+  --active-interval 120 \
+  --dormancy-threshold 600
+```
+
+---
+
+### `config get/set` — Read/write individual config keys
+
+Direct access to vault config via dot-path notation. Values are parsed as JSON first, then fall back to raw strings.
+
+**Usage:**
+
+```sh
+# Read a value
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js config get intelligence.llm.model
+
+# Write a value
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js config set intelligence.llm.model qwen2.5-coder:14b
+
+# Write a non-string value (parsed as JSON)
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js config set digest.enabled true
+```
+
+Restart the daemon after changes: `node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js restart`
+
+---
+
+## Diagnostic Commands
+
+### `detect-providers` — Probe available LLM providers
+
+Checks Ollama, LM Studio, and Anthropic for availability and lists discovered models.
+
+No flags.
+
+Output is JSON:
+
+```json
+{
+  "ollama": { "available": true, "models": ["qwen2.5-coder:14b", "nomic-embed-text"] },
+  "lm-studio": { "available": false, "models": [] },
+  "anthropic": { "available": true, "models": [] }
+}
+```
+
+---
+
+### `verify` — Test LLM and embedding connectivity
+
+Sends a test prompt to the LLM and a test embed to the embedding provider. Exits 0 if both pass, 1 if either fails.
+
+No flags.
+
+```sh
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js verify
+```
+
+---
+
+### `stats` — Vault health and daemon status
+
+Shows session/spore/plan counts, spore type breakdown, vector count, and daemon PID/port.
+
+No flags.
+
+```sh
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js stats
+```
+
+Typical output:
+
+```
+Sessions:  12
+Spores:    47  (decision: 15, gotcha: 12, trade_off: 8, discovery: 7, bug_fix: 5)
+Plans:     2
+Vectors:   61
+
+Daemon:    PID 38291 on port 60942 (2 active sessions)
+```
+
+---
+
+### `logs` — Tail, filter, and follow daemon logs
+
+| Flag | Type | Description |
+|------|------|-------------|
+| `-f` / `--follow` | boolean | Watch for new log entries (blocks until Ctrl+C) |
+| `-n` / `--tail <number>` | number | Number of lines to show (default: 100) |
+| `-l` / `--level <level>` | string | Filter by level: `debug`, `info`, `warning`, `error` |
+| `-c` / `--component <name>` | string | Filter by component |
+| `--since <timestamp>` | string | Show logs after ISO timestamp |
+| `--until <timestamp>` | string | Show logs before ISO timestamp |
+
+Components: `processor`, `embeddings`, `hooks`, `lifecycle`, `daemon`, `lineage`, `watcher`
+
+**Examples:**
+
+```sh
+# Show last 20 lines
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js logs -n 20
+
+# Follow errors from the processor
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js logs -f -l error -c processor
+
+# Show logs from a specific window
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js logs \
+  --since 2025-01-01T10:00:00Z \
+  --until 2025-01-01T11:00:00Z
+```
+
+---
+
+## Query Commands
+
+### `search <query>` — Combined semantic + FTS search
+
+Runs semantic search (primary) with FTS fallback across sessions, spores, and plans.
+
+```sh
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js search "why did we choose sqlite over postgres"
+```
+
+---
+
+### `vectors <query>` — Raw vector similarity scores
+
+Shows all results with similarity scores and no threshold filtering. Useful for tuning embedding thresholds.
+
+```sh
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js vectors "session lifecycle hooks"
+```
+
+---
+
+### `session [id|latest]` — Display a session note
+
+| Argument | Description |
+|----------|-------------|
+| `latest` | Show the most recent session (default if omitted) |
+| `<id>` | Session ID substring — matches the first session containing the substring |
+
+```sh
+# Show latest session
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js session
+
+# Show a specific session by partial ID
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js session ac5220
+```
+
+---
+
+## Maintenance Commands
+
+### `restart` — Kill and respawn the daemon
+
+Sends SIGTERM to the running daemon, waits for it to exit, and spawns a fresh instance with a health check.
+
+No flags.
+
+```sh
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js restart
+```
+
+Run this after any daemon code changes to pick up new behavior.
+
+---
+
+### `rebuild` — Full FTS + vector reindex
+
+Re-indexes all vault notes. Superseded and archived spores are skipped.
+
+No flags.
+
+```sh
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js rebuild
+```
+
+Run this after changing the embedding model (via `setup-llm`) to regenerate all embeddings with the new model.
+
+---
+
+### `reprocess` — Re-extract observations from transcripts
+
+Re-reads session transcripts, re-extracts observations with the current LLM, and re-indexes. Existing spores are preserved — new extractions are additive.
+
+| Flag | Type | Description |
+|------|------|-------------|
+| `--session <id>` | string | Session ID substring — reprocess only matching sessions |
+| `--index-only` | boolean | Skip LLM extraction, FTS-only reindex |
+
+**Examples:**
+
+```sh
+# Reprocess all sessions
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js reprocess
+
+# Reprocess one session
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js reprocess --session ac5220
+
+# Re-index without re-extracting
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js reprocess --index-only
+```
+
+---
+
+## Info Commands
+
+### `version` — Show plugin version
+
+```sh
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js version
+```
+
+Also available as `--version` or `-v`.

package/skills/myco/references/vault-status.md

@@ -0,0 +1,224 @@
+# Vault Health Reference
+
+This reference guides vault health assessment. Use it whenever you need to inspect the state of the Myco vault — whether debugging a problem, responding to a user's question about their vault, or running a routine check.
+
+**Primary data source:** `node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js stats`
+
+Supplement with `logs` for recent daemon activity and direct file checks for digest status.
+
+---
+
+## Running Stats
+
+Run the stats command to get a snapshot of vault state:
+
+```sh
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js stats
+```
+
+The output includes:
+
+- **Session/spore/plan counts** — total notes in each category
+- **Spore type breakdown** — counts per observation type (decision, gotcha, trade_off, discovery, bug_fix, etc.)
+- **Vector count** — number of indexed embeddings
+- **Daemon PID/port/active sessions** — runtime state
+
+Typical output looks like:
+
+```
+Sessions:  12
+Spores:    47  (decision: 15, gotcha: 12, trade_off: 8, discovery: 7, bug_fix: 5)
+Plans:     2
+Vectors:   61
+
+Daemon:    PID 38291 on port 60942 (2 active sessions)
+```
+
+If the command fails or returns no output, the daemon may not be running or the vault may not be configured. Check daemon status next.
+
+---
+
+## Daemon Health
+
+The daemon runs as a background HTTP process. Its state is persisted in `<vault>/daemon.json`.
+
+**To check daemon status:**
+
+1. Read `<vault>/daemon.json` for the PID and port.
+2. Verify the PID is alive: `kill -0 <pid>` returns 0 if running.
+3. HTTP health check: `curl http://localhost:<port>/health`
+
+**If the daemon is not running:**
+
+> "Daemon not running. It will start automatically on next session."
+
+You can also restart it manually:
+
+```sh
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js restart
+```
+
+Do not treat a stopped daemon as a vault error — it is normal between sessions. The daemon starts on demand when a session begins.
+
+---
+
+## Digest Status
+
+The digest system synthesizes vault knowledge into pre-computed context extracts. Check its state with direct file inspection.
+
+**Is digest enabled?**
+
+Read `digest.enabled` from `<vault>/myco.yaml`. If false or absent, skip digest checks.
+
+**Which extracts exist?**
+
+List files in `<vault>/digest/`:
+
+- `extract-1500.md` — ~1500 token tier
+- `extract-3000.md` — ~3000 token tier
+- `extract-5000.md` — ~5000 token tier
+- `extract-10000.md` — ~10000 token tier
+
+Report each file's size and last-modified timestamp. Missing tiers mean the digest has not yet run for that tier, or the tier is not configured.
+
+**Last digest cycle:**
+
+Read the last line of `<vault>/digest/trace.jsonl`. Each line is a JSON object with:
+
+- `cycleId` — short identifier for the cycle
+- `timestamp` — ISO timestamp of when the cycle ran
+- `tiersGenerated` — which tiers were written
+- `substrateCount` — number of vault notes processed as input
+- `durationMs` — how long the cycle took
+
+**Metabolism config:**
+
+From `myco.yaml`, report:
+
+- Configured tiers (e.g., `[1500, 3000, 5000, 10000]`)
+- Inject tier — which tier is injected at session start
+- Context window size
+- Digest model — if `digest.intelligence.model` is set, report it; otherwise note "inherits from main LLM"
+
+---
+
+## Intelligence Backend Health
+
+Test connectivity to the configured LLM and embedding providers:
+
+```sh
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js verify
+```
+
+This reports whether the LLM provider and embedding provider are reachable. If either is unreachable:
+
+- Check that the provider process is running (Ollama, LM Studio, etc.)
+- Review the model name in `myco.yaml` for typos
+- Re-run configuration with the appropriate CLI setup command
+
+---
+
+## Issue Detection
+
+Check for these problems when assessing vault health:
+
+| Issue | How to detect | Meaning |
+|-------|--------------|---------|
+| **Stale buffers** | `.jsonl` files in `<vault>/buffer/` older than 24 hours | Events were captured but never processed — LLM may have been unavailable |
+| **Missing index** | `<vault>/index.db` does not exist | FTS search will not work; suggest `node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js rebuild` |
+| **Missing vectors** | `<vault>/vectors.db` does not exist | Semantic search disabled; embeddings may be unconfigured |
+| **Old config version** | `version` in `myco.yaml` is less than `2` | Vault may need migration; suggest running `/myco-setup` |
+
+Report all issues found, or "None found." if the vault is clean.
+
+---
+
+## Recent Activity
+
+View recent daemon log output:
+
+```sh
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js logs -n 20
+```
+
+This shows the last 20 lines of daemon logs — useful for spotting errors, slow processing, or repeated failures.
+
+Also report the last 3 sessions with:
+
+- Session ID (short form)
+- Title
+- Started/ended timestamps
+- Number of spores extracted
+- Parent session, if lineage was detected
+
+---
+
+## Report Format
+
+Present vault health as a structured report. Always show all sections; use "None" rather than omitting sections with no data.
+
+```
+=== Myco Vault ===
+Path: <vault-path>
+Config: v<version>
+
+--- Intelligence ---
+LLM:       <provider> / <model> (<reachable|unreachable>)
+Embedding: <provider> / <model> (<reachable|unreachable>)
+
+--- Daemon ---
+PID:      <pid> (<running|stopped>)
+Port:     <port>
+Sessions: <count> active
+
+--- Vault ---
+Sessions:  <count>
+Spores:    <total> (<breakdown by type>)
+Plans:     <count>
+Vectors:   <count>
+
+--- Digest ---
+Enabled:    <yes|no>
+Tiers:      <list>
+Inject:     <tier>
+Model:      <model> (<inherited|configured>)
+Last cycle: <id> (<time ago>, <tiers>, <notes>, <duration>)
+Extracts:   <tier> (<size>), ...
+
+--- Lineage ---
+Links: <count> (<breakdown by type>)
+
+--- Recent Sessions ---
+1. [<id>] "<title>" (<duration>, <spore count> spores)
+2. [<id>] "<title>" (<duration>, <spore count> spores, parent: <id>)
+3. [<id>] "<title>" (<duration>, <spore count> spores)
+
+--- Issues ---
+<issues or "None found.">
+```
+
+Adapt to what is actually available. If the daemon is stopped, omit PID/port/sessions. If digest is disabled, abbreviate that section. Keep the structure recognizable.
+
+---
+
+## Lineage
+
+If `<vault>/lineage.json` exists, report the link count and breakdown:
+
+- **clear** — session explicitly resumed a prior session
+- **inferred** — heuristic match based on timing or working directory
+- **semantic_similarity** — matched by content similarity of opening context
+
+Lineage tracks parent-child relationships between sessions. A high count of `semantic_similarity` links (relative to `clear`) may indicate sessions that should be explicitly resumed rather than started fresh.
+
+---
+
+## Suggested Actions
+
+| Issue | Action |
+|-------|--------|
+| Daemon not running | Will auto-start on next session; or run `node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js restart` |
+| Stale buffers | Check if LLM provider was down during those sessions; events will process on next daemon start |
+| Missing index | Run `node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js rebuild` to regenerate FTS and vector indexes |
+| Provider unreachable | Ensure the provider is running (e.g., `ollama serve`); verify model name in `myco.yaml`; reconfigure with CLI commands |
+| Config version < 2 | Run `/myco-setup` to migrate the vault configuration |

package/skills/setup/SKILL.md

@@ -0,0 +1,146 @@
+---
+name: setup
+description: >-
+  Initialize Myco in a new project — guided first-time setup for vault,
+  LLM provider, and intelligence backend
+user-invocable: true
+allowed-tools: Bash, AskUserQuestion, Skill
+---
+
+# Setup — Guided Myco Onboarding
+
+This skill guides a first-time Myco setup from zero to a working vault. It detects the system's hardware and available providers, asks one question at a time, and delegates all configuration to CLI commands — never touching config files directly. If a vault already exists, it hands off to the `myco` skill for reconfiguration or status.
+
+## Step 1: Check Vault Existence
+
+Run:
+
+```bash
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js stats
+```
+
+- If the command **succeeds** (exit code 0): tell the user "Myco is already configured" and invoke the `myco` skill to handle reconfiguration or display status. Stop here — do not continue with the setup flow.
+- If the command **fails** (exit code non-zero or vault not found): proceed to Step 2.
+
+## Step 2: Detect System
+
+Run both commands in parallel:
+
+```bash
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js detect-providers
+```
+
+Parse the JSON output. The result will list providers and their `available` field (boolean) and available models. Keep this data — you will use it in Step 3.
+
+For RAM detection, run the appropriate command for the OS:
+
+- **macOS:** `sysctl -n hw.memsize` — result is bytes; divide by `1073741824` (1024³) to get GB
+- **Linux:** parse `/proc/meminfo` for the `MemTotal` line — result is in kB; divide by `1048576` to get GB
+
+Use the RAM value to determine the recommended tier from `references/model-recommendations.md`:
+
+| RAM | Recommended Intelligence Model | Digest Context Window | Default Inject Tier |
+|-----|--------------------------------|-----------------------|---------------------|
+| 64GB+ | `qwen3.5:35b` | 65536 | 3000 |
+| 32–64GB | `qwen3.5:27b` | 32768 | 3000 |
+| 16–32GB | `qwen3.5:latest` (~10B) | 16384 | 1500 |
+| 8–16GB | `qwen3.5:4b` | 8192 | 1500 |
+
+Record: detected RAM (GB), recommended model, digest context window, and default inject tier. You will use these as defaults in the questions below.
+
+## Step 3: Ask Questions
+
+Ask one question at a time. Do not batch questions. Wait for each answer before asking the next.
+
+### Question 1: Vault Location
+
+Ask the user where to store the vault. Present three choices:
+
+- **Project-local** — `.myco/` in the current directory
+- **Centralized** — `~/.myco/vaults/<project-name>/` (where `<project-name>` is the current directory's basename)
+- **Custom path** — the user types a path
+
+Record the resolved vault path.
+
+### Question 2: Provider and Model
+
+From the `detect-providers` output, list only providers where `available` is `true`. Present them as choices. For each available provider, list its available models.
+
+Pre-select the recommended model from Step 2 if it appears in the list. If the recommended model is not installed:
+
+- **Ollama:** offer to run `ollama pull <recommended-model>` before continuing. Ask "Pull now or choose a different model?"
+- **LM Studio:** tell the user to open LM Studio, search for `<recommended-model>`, and download it. Offer to wait or to let the user choose a different available model.
+
+Record the chosen provider and model.
+
+### Question 3: Embedding Model
+
+List embedding models from the chosen provider. Exclude Anthropic — it does not support embeddings.
+
+If no embedding models are installed:
+
+- **Ollama:** offer to run `ollama pull bge-m3`. If the user accepts, run it before continuing.
+- **LM Studio:** tell the user to search for and download an embedding model (suggest `bge-m3` or any model with `text-embedding` in the name).
+
+Recommend `bge-m3` as the default. Record the chosen embedding provider and model.
+
+### Question 4: Inject Tier
+
+Show the inject tier options appropriate for the detected RAM, with the default pre-selected:
+
+| Tier | Description |
+|------|-------------|
+| 1500 | Executive briefing — fastest, lightest |
+| 3000 | Team standup — recommended for most setups |
+| 5000 | Deep onboarding |
+| 10000 | Institutional knowledge — richest context |
+
+Show only the tiers available for the user's RAM tier (per the table in Step 2). Pre-select the default. Tell the user: "Agents can always request a different tier on-demand via the `myco_context` MCP tool."
+
+Record the chosen inject tier.
+
+## Step 4: Execute
+
+Run the following commands in sequence, substituting the recorded values. Show each command before running it.
+
+```bash
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js init \
+  --vault <chosen-vault-path> \
+  --llm-provider <provider> \
+  --llm-model <model> \
+  --embedding-provider <embedding-provider> \
+  --embedding-model <embedding-model>
+```
+
+```bash
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js setup-digest \
+  --context-window <digest-context-window-from-ram-table> \
+  --inject-tier <chosen-inject-tier>
+```
+
+```bash
+node ${CLAUDE_PLUGIN_ROOT}/dist/src/cli.js verify
+```
+
+If any command fails, report the error and stop. Do not continue to the next command on failure. Show the full error output to the user and ask how to proceed.
+
+## Step 5: Report
+
+Display a summary table:
+
+| Setting | Value |
+|---------|-------|
+| Vault path | `<resolved path>` |
+| Provider | `<provider>` / `<model>` |
+| Embedding | `<embedding-provider>` / `<embedding-model>` |
+| Digest | enabled (context: `<context-window>`, inject: `<inject-tier>`) |
+| RAM detected | `<X>` GB |
+
+Tell the user: "Myco is ready. Start a new session to begin capturing knowledge."
+
+## Constraints
+
+- All writes via CLI commands — never read or modify `myco.yaml` directly.
+- All provider detection via `detect-providers` — no raw HTTP calls to provider APIs.
+- One question at a time — do not batch questions or present them together.
+- Two model choices in guided setup: intelligence model and embedding model. For a separate dedicated digestion model, direct the user to run `setup-digest` with a `--provider` and `--model` flag after setup completes (see `references/model-recommendations.md` Advanced section).