pi update

pi/.pi/agent/agents/explorer.md

@@ -2,7 +2,7 @@
 name: explorer
 description: Comprehensive codebase and knowledge-base explorer. Maps architecture, traces dependencies, synthesizes cross-cutting context with full code snippets and rationale. Use for deep refactoring, architectural decisions, or understanding complex subsystems. Do NOT use when the user has already provided explicit file paths or when a direct file read would suffice — only invoke for open-ended exploration where the relevant files are unknown.
 tools: read, bash, write, mcp:qmd, mcp:opty
-model: opencode/big-pickle
+model: opencode-go/mimo-v2-pro
 defaultProgress: true
 ---
 

pi/.pi/agent/agents/planner.md

@@ -2,7 +2,7 @@
 name: planner
 description: Creates detailed implementation plans from scout context and task requirements. Writes the plan to a file for the next agent in the chain.
 tools: read, write, grep, find, ls
-model: opencode/qwen3.6-plus-free
+model: anthropic/claude-opus-4-6
 output: plan.md
 defaultReads: scout.md
 ---

pi/.pi/agent/agents/scout.md

@@ -1,7 +1,8 @@
 ---
 name: scout
 description: Fast codebase recon. Finds relevant files, types, and patterns, then writes structured context for the next agent in the chain.
-tools: read, write, grep, find, ls, bash, mcp:opty, mcp:qmd
+tools: read, write, grep, find, ls, bash
+skills: opty, qmd
 model: opencode-go/mimo-v2-pro
 output: scout.md
 ---
@@ -21,17 +22,17 @@ When your task contains `[Write to: path]`, write your COMPLETE report to that e
 Without `[Write to:]`, output your full report as text.
 
 ## Tools
-- **opty** — semantic/HDC code search: use `opty_opty_query` to find functions, types, imports by meaning; `opty_opty_ast` for a structural overview of a file or the project
-- **qmd** — knowledge base search: use `qmd_query` to find docs, notes, or prior context by keyword or vector; `qmd_get` / `qmd_multi_get` to retrieve full documents
-- **grep/find/bash** — fallback for exact patterns, file discovery, or anything the semantic tools miss
+- **opty** — semantic/HDC code search via CLI: `opty query "description"` to find functions/types/imports by meaning; output includes file + line number
+- **qmd** — knowledge base search via CLI: `qmd query $'lex: X\nvec: Y'` to find docs/notes by keyword or vector; `qmd get <path>` / `qmd multi-get <glob>` to retrieve full documents
+- **grep/find/bash** — for exact patterns, file discovery, or anything the semantic tools miss
 - **read** — read specific file sections once you know where to look
 
 ## Strategy
-1. Use `opty_opty_query` to semantically locate relevant functions/types (fast, no file reading needed)
-2. Use `qmd_query` to check if there's relevant documentation or prior context in the knowledge base
+1. Use `opty query "..."` to semantically locate relevant functions/types (fast, no file reading needed)
+2. Use `qmd query "..."` to check if there's relevant documentation or prior context in the knowledge base
 3. grep/find for exact patterns or when semantic search isn't precise enough
 4. Read key sections (not entire files — target the relevant functions/types)
-5. Use `opty_opty_ast` on key files to quickly understand their structure without reading everything
+5. For file structure overview, use grep for exports/types or read the top ~50 lines of a file
 6. Identify types, interfaces, key functions
 7. Note dependencies between files
 8. Flag anything surprising or risky

pi/.pi/agent/extensions/claude-agent-sdk/package.json

@@ -0,0 +1,11 @@
+{
+  "name": "claude-agent-sdk",
+  "type": "module",
+  "dependencies": {
+    "@anthropic-ai/claude-agent-sdk": "^0.2.92",
+    "change-case": "^5.4.4"
+  },
+  "pi": {
+    "extensions": ["./index.ts"]
+  }
+}

pi/.pi/agent/extensions/exa-mcp.json

@@ -0,0 +1,12 @@
+{
+  "url": "https://mcp.exa.ai/mcp",
+  "apiKey": null,
+  "tools": [
+    "web_search_exa",
+    "get_code_context_exa"
+  ],
+  "timeoutMs": 30000,
+  "protocolVersion": "2025-06-18",
+  "maxBytes": 51200,
+  "maxLines": 2000
+}

pi/.pi/agent/extensions/usage-bars/core.ts

@@ -36,8 +36,8 @@ export function writeUsageCache(cache: UsageCache): void {
   } catch {}
 }
 
-export type ProviderKey = "codex" | "claude" | "zai" | "gemini" | "antigravity";
-export type OAuthProviderId = "openai-codex" | "anthropic" | "google-gemini-cli" | "google-antigravity";
+export type ProviderKey = "codex" | "claude" | "zai" | "gemini" | "antigravity" | "opencode-go";
+export type OAuthProviderId = "openai-codex" | "anthropic" | "google-gemini-cli" | "google-antigravity" | "opencode-go";
 
 export interface AuthData {
   "openai-codex"?: { access?: string; refresh?: string; expires?: number };
@@ -45,6 +45,7 @@ export interface AuthData {
   zai?: { key?: string; access?: string; refresh?: string; expires?: number };
   "google-gemini-cli"?: { access?: string; refresh?: string; projectId?: string; expires?: number };
   "google-antigravity"?: { access?: string; refresh?: string; projectId?: string; expires?: number };
+  "opencode-go"?: { key?: string; access?: string };
 }
 
 export interface UsageData {
@@ -61,6 +62,127 @@ export interface UsageData {
 
 export type UsageByProvider = Record<ProviderKey, UsageData | null>;
 
+// OpenCode Go usage tracking (local, since no API exists yet)
+export interface OpenCodeGoLocalUsage {
+  /** Dollar value used in the 5-hour window */
+  fiveHourUsed: number;
+  /** Dollar value used in the weekly window */
+  weeklyUsed: number;
+  /** Dollar value used in the monthly window */
+  monthlyUsed: number;
+  /** Timestamp of the last update */
+  lastUpdated: number;
+  /** When the current 5-hour window started */
+  fiveHourWindowStart: number;
+  /** When the current week started (Unix ms) */
+  weekStart: number;
+  /** When the current month started (Unix ms) */
+  monthStart: number;
+}
+
+const OPENCODE_GO_USAGE_FILE = path.join(os.homedir(), ".pi", "agent", "opencode-go-usage.json");
+const OPENCODE_GO_FIVE_HOUR_LIMIT = 12;
+const OPENCODE_GO_WEEKLY_LIMIT = 30;
+const OPENCODE_GO_MONTHLY_LIMIT = 60;
+const OPENCODE_GO_FIVE_HOUR_MS = 5 * 60 * 60 * 1000;
+
+export function readOpenCodeGoUsage(): OpenCodeGoLocalUsage | null {
+  try {
+    const raw = fs.readFileSync(OPENCODE_GO_USAGE_FILE, "utf-8");
+    const parsed = JSON.parse(raw);
+    if (typeof parsed?.lastUpdated === "number") return parsed as OpenCodeGoLocalUsage;
+  } catch {}
+  return null;
+}
+
+export function writeOpenCodeGoUsage(usage: OpenCodeGoLocalUsage): void {
+  try {
+    const dir = path.dirname(OPENCODE_GO_USAGE_FILE);
+    if (!fs.existsSync(dir)) fs.mkdirSync(dir, { recursive: true });
+    const tmp = `${OPENCODE_GO_USAGE_FILE}.tmp-${process.pid}-${Date.now()}`;
+    fs.writeFileSync(tmp, JSON.stringify(usage, null, 2));
+    fs.renameSync(tmp, OPENCODE_GO_USAGE_FILE);
+  } catch {}
+}
+
+export function resetOpenCodeGoUsageIfNeeded(existing: OpenCodeGoLocalUsage | null): OpenCodeGoLocalUsage {
+  const now = Date.now();
+  const nowDate = new Date(now);
+
+  // Start with defaults
+  let usage: OpenCodeGoLocalUsage = existing ?? {
+    fiveHourUsed: 0,
+    weeklyUsed: 0,
+    monthlyUsed: 0,
+    lastUpdated: now,
+    fiveHourWindowStart: now,
+    weekStart: now,
+    monthStart: now,
+  };
+
+  // Reset 5-hour window if expired
+  if (now - usage.fiveHourWindowStart >= OPENCODE_GO_FIVE_HOUR_MS) {
+    usage.fiveHourUsed = 0;
+    usage.fiveHourWindowStart = now;
+  }
+
+  // Reset weekly window (Monday-based)
+  const dayOfWeek = nowDate.getDay();
+  const daysSinceMonday = dayOfWeek === 0 ? 6 : dayOfWeek - 1;
+  const thisMonday = new Date(nowDate);
+  thisMonday.setDate(nowDate.getDate() - daysSinceMonday);
+  thisMonday.setHours(0, 0, 0, 0);
+  if (usage.weekStart < thisMonday.getTime()) {
+    usage.weeklyUsed = 0;
+    usage.weekStart = thisMonday.getTime();
+  }
+
+  // Reset monthly window (1st of month)
+  const thisMonthStart = new Date(nowDate.getFullYear(), nowDate.getMonth(), 1);
+  if (usage.monthStart < thisMonthStart.getTime()) {
+    usage.monthlyUsed = 0;
+    usage.monthStart = thisMonthStart.getTime();
+  }
+
+  usage.lastUpdated = now;
+  return usage;
+}
+
+export function addOpenCodeGoSpend(dollars: number): void {
+  let usage = resetOpenCodeGoUsageIfNeeded(readOpenCodeGoUsage());
+  usage.fiveHourUsed += dollars;
+  usage.weeklyUsed += dollars;
+  usage.monthlyUsed += dollars;
+  usage.lastUpdated = Date.now();
+  writeOpenCodeGoUsage(usage);
+}
+
+export function getOpenCodeGoUsageData(): UsageData {
+  const usage = resetOpenCodeGoUsageIfNeeded(readOpenCodeGoUsage());
+  if (!usage) {
+    return { session: 0, weekly: 0, error: "no local usage data" };
+  }
+
+  const sessionPct = Math.min(100, (usage.fiveHourUsed / OPENCODE_GO_FIVE_HOUR_LIMIT) * 100);
+  const weeklyPct = Math.min(100, (usage.weeklyUsed / OPENCODE_GO_WEEKLY_LIMIT) * 100);
+
+  // Calculate resets
+  const fiveHourEnd = usage.fiveHourWindowStart + OPENCODE_GO_FIVE_HOUR_MS;
+  const fiveHourRemaining = Math.max(0, fiveHourEnd - Date.now());
+
+  const weekEnd = usage.weekStart + 7 * 24 * 60 * 60 * 1000;
+  const weekRemaining = Math.max(0, weekEnd - Date.now());
+
+  return {
+    session: sessionPct,
+    weekly: weeklyPct,
+    sessionResetsIn: formatDuration(Math.round(fiveHourRemaining / 1000)),
+    weeklyResetsIn: formatDuration(Math.round(weekRemaining / 1000)),
+    extraSpend: usage.monthlyUsed,
+    extraLimit: OPENCODE_GO_MONTHLY_LIMIT,
+  };
+}
+
 export interface UsageEndpoints {
   zai: string;
   gemini: string;
@@ -600,12 +722,14 @@ export function detectProvider(
   if (typeof model === "string") return null;
 
   const provider = (model.provider || "").toLowerCase();
+  const id = (model.id || "").toLowerCase();
 
   if (provider === "openai-codex") return "codex";
   if (provider === "anthropic") return "claude";
   if (provider === "zai") return "zai";
   if (provider === "google-gemini-cli") return "gemini";
   if (provider === "google-antigravity") return "antigravity";
+  if (provider === "opencode-go" || id.startsWith("opencode-go/")) return "opencode-go";
 
   return null;
 }
@@ -615,6 +739,7 @@ export function providerToOAuthProviderId(active: ProviderKey | null): OAuthProv
   if (active === "claude") return "anthropic";
   if (active === "gemini") return "google-gemini-cli";
   if (active === "antigravity") return "google-antigravity";
+  if (active === "opencode-go") return "opencode-go";
   return null;
 }
 
@@ -629,6 +754,9 @@ export function canShowForProvider(active: ProviderKey | null, auth: AuthData |
   if (active === "antigravity") {
     return !!(auth["google-antigravity"]?.access || auth["google-antigravity"]?.refresh) && !!endpoints.antigravity;
   }
+  if (active === "opencode-go") {
+    return !!(auth["opencode-go"]?.key || auth["opencode-go"]?.access);
+  }
   return false;
 }
 
@@ -654,6 +782,7 @@ export async function fetchAllUsages(config: FetchAllUsagesConfig = {}): Promise
     zai: null,
     gemini: null,
     antigravity: null,
+    "opencode-go": null,
   };
 
   if (!auth) return results;
@@ -733,6 +862,11 @@ export async function fetchAllUsages(config: FetchAllUsagesConfig = {}): Promise
     }
   }
 
+  // OpenCode Go uses local tracking (no public API yet)
+  if (authData["opencode-go"]?.key || authData["opencode-go"]?.access) {
+    results["opencode-go"] = getOpenCodeGoUsageData();
+  }
+
   await Promise.all(tasks);
   return results;
 }

pi/.pi/agent/extensions/usage-bars/index.ts

@@ -79,6 +79,7 @@ const PROVIDER_LABELS: Record<ProviderKey, string> = {
   zai: "Z.AI",
   gemini: "Gemini",
   antigravity: "Antigravity",
+  "opencode-go": "Go",
 };
 
 // ---------------------------------------------------------------------------
@@ -163,6 +164,7 @@ class UsageSelectorComponent extends Container implements Focusable {
       { key: "zai", name: "Z.AI" },
       { key: "gemini", name: "Gemini" },
       { key: "antigravity", name: "Antigravity" },
+      { key: "opencode-go", name: "Go" },
     ];
     this.allItems = [];
     for (const p of providers) {
@@ -287,7 +289,7 @@ interface PollOptions {
 export default function (pi: ExtensionAPI) {
   const endpoints = resolveUsageEndpoints();
   const state: UsageState = {
-    codex: null, claude: null, zai: null, gemini: null, antigravity: null,
+    codex: null, claude: null, zai: null, gemini: null, antigravity: null, "opencode-go": null,
     lastPoll: 0, activeProvider: null,
   };
 
@@ -433,6 +435,10 @@ export default function (pi: ExtensionAPI) {
       result = creds?.access
         ? await fetchGoogleUsage(creds.access, endpoints.gemini, creds.projectId, "gemini", { endpoints })
         : { session: 0, weekly: 0, error: "missing access token (try /login again)" };
+    } else if (active === "opencode-go") {
+      // OpenCode Go uses local tracking (no public usage API yet)
+      const { getOpenCodeGoUsageData } = await import("./core");
+      result = getOpenCodeGoUsageData();
     } else {
       const creds = effectiveAuth["google-antigravity"];
       result = creds?.access
@@ -561,6 +567,22 @@ export default function (pi: ExtensionAPI) {
     void poll({ forceFresh: true });
   });
 
+  // Listen for OpenCode Go spend events from other extensions
+  pi.events.on("opencode-go:spend", async (amount: number) => {
+    if (typeof amount === "number" && amount > 0) {
+      const { addOpenCodeGoSpend } = await import("./core");
+      addOpenCodeGoSpend(amount);
+      // Invalidate cache and re-poll
+      const cache = readUsageCache();
+      if (cache?.data?.["opencode-go"]) {
+        const nextCache: import("./core").UsageCache = { ...cache, data: { ...cache.data } };
+        delete nextCache.data["opencode-go"];
+        writeUsageCache(nextCache);
+      }
+      void poll({ forceFresh: true });
+    }
+  });
+
   // ── /usage command ───────────────────────────────────────────────────────
 
   pi.registerCommand("usage", {

pi/.pi/agent/models.json

@@ -23,6 +23,15 @@
           "maxTokens": 131072,
           "cost": { "input": 1.0, "output": 3.2, "cacheRead": 0.2, "cacheWrite": 0 }
         },
+        {
+          "id": "glm-5.1",
+          "name": "GLM-5.1",
+          "reasoning": true,
+          "input": ["text"],
+          "contextWindow": 204800,
+          "maxTokens": 131072,
+          "cost": { "input": 1.0, "output": 3.2, "cacheRead": 0.2, "cacheWrite": 0 }
+        },
         {
           "id": "kimi-k2.5",
           "name": "Kimi K2.5",

pi/.pi/agent/prompts/implement-critical.md

@@ -1,5 +1,5 @@
 ---
-description: "Maximum quality pipeline — deep scout, Max planning, Opus plan review, approval gate, Sonnet coding, Opus code review"
+description: "Maximum quality pipeline — deep scout, thorough planning, plan review, approval gate, coding, code review"
 ---
 
 Use the subagent tool to implement with maximum quality. This is for high-risk or architecturally significant changes.
@@ -30,7 +30,7 @@ If the user requests changes, revise the plan (re-run planner with the feedback)
 - For each coder run, include the approved plan verbatim: "Implement the following plan step(s). Do NOT deviate.\n\n<plan>\n{the approved plan steps}\n</plan>"
 - For multiple independent steps, run them in parallel using separate coder-claude tasks, each assigned to specific files/plan steps to avoid conflicts
 
-## Step 4: Opus review
+## Step 4: Code review
 Run the "reviewer" agent on all changes with this task: "Review all changes made for: $@\n\nCheck for correctness, edge cases, error handling, type safety, and consistency with the approved plan."
 
 ## Step 5: Fix

pi/.pi/agent/prompts/implement.md

@@ -29,7 +29,7 @@ If the user requests changes to the plan, revise and present again before implem
 ## Step 3: Implement
 
 Once approved:
-- Execute the plan steps using "coder" for sequential steps, or "coder-parallel" with parallel tasks if the plan identified parallelizable steps
+- Execute the plan steps using the "coder" agent
 - When running coder, always wrap the plan step(s) in the task: "Implement the following plan step(s). Do NOT deviate.\n\n<plan>\n{the approved plan steps}\n</plan>"
 - Run the "reviewer" agent on all changes
 - If NEEDS_FIXES, run the "fixer" agent with the review output
@@ -43,26 +43,13 @@ After the final step, summarize: what was done, what files changed, what was rev
 - **NEVER skip the approval gate**. Always present the plan and wait.
 - Always pass scout context forward using {previous} in chain mode — this is how the planner and plan-reviewer receive the scout's findings.
 - When running the coder, always include the approved plan verbatim in the task so the coder has full context.
-- For parallel coder tasks, clearly assign each coder to specific files/plan steps to avoid conflicts.
 
 ## Agent Failure and Fallback
 
 When a subagent returns empty output or an error (rate limit, credit exhaustion, connection failure):
 
 1. **Retry once** with the same agent and model — transient failures are common.
-2. **If still failing, retry with the cross-family fallback model** using the `model` override parameter:
-
-| Agent | Primary model | Fallback model |
-|-------|--------------|----------------|
-| scout | zai/glm-4.7-flash | anthropic/claude-haiku-4-5 |
-| deep-scout, coder-parallel | zai/glm-4.7 | anthropic/claude-sonnet-4-6 |
-| planner, coder | zai/glm-5.1 | anthropic/claude-opus-4-6 |
-| reviewer-quick, explorer | anthropic/claude-haiku-4-5 | zai/glm-4.7-flash |
-| coder-claude, fixer | anthropic/claude-sonnet-4-6 | zai/glm-5.1 |
-| plan-reviewer, reviewer | anthropic/claude-opus-4-6 | zai/glm-5.1 |
-
-Example: `subagent({ agent: "scout", task: "...", model: "anthropic/claude-haiku-4-5" })`
-
+2. **If still failing, retry with the cross-family fallback model** using the `model` override parameter. See the fallback table in the `subagent-implement` SKILL.md for the current primary/fallback mapping.
 3. **If the fallback also fails**, do the work yourself (read the relevant files directly and produce the scout/plan/review output inline). Inform the user which agent failed, what error was returned, and what you did instead.
 
 Do NOT silently absorb failures. Always surface them to the user even when working around them.

pi/.pi/agent/prompts/plan.md

@@ -1,5 +1,5 @@
 ---
-description: "Scout + plan + Opus plan review — no implementation"
+description: "Scout + plan + plan review — no implementation"
 ---
 
 Use the subagent tool with a chain to plan (but NOT implement) the following:
@@ -19,8 +19,5 @@ Present the plan and the review to me. Do NOT proceed to implementation.
 If any agent returns empty output or an error (rate limit, credit exhaustion, connection failure):
 
 1. Retry once with the same agent.
-2. If still failing, retry with the cross-family fallback using the `model` override:
-   - scout (zai/glm-4.7-flash fails) → `model: "anthropic/claude-haiku-4-5"`
-   - planner (zai/glm-5.1 fails) → `model: "anthropic/claude-opus-4-6"`
-   - plan-reviewer (anthropic/claude-opus-4-6 fails) → `model: "zai/glm-5.1"`
+2. If still failing, retry with the cross-family fallback using the `model` override. See the fallback table in the `subagent-plan` SKILL.md for the current mapping.
 3. If the fallback also fails, do the work yourself and tell me which agent failed and why.

pi/.pi/agent/skills/opty/SKILL.md

@@ -0,0 +1,37 @@
+---
+name: opty
+description: Semantic code search using HDC (Hyperdimensional Computing). Finds functions, types, imports, and variables in the indexed codebase by meaning. Use when searching for code by concept — e.g. "error handling functions", "database types", "authentication flow".
+compatibility: Requires opty CLI and a running daemon. Install from source or binary.
+allowed-tools: Bash(opty:*)
+---
+
+# Opty — HDC Code Search
+
+Semantic code search via Hyperdimensional Computing. Indexes the codebase and finds relevant functions, types, and imports by meaning rather than exact text. The daemon auto-starts when you run any opty command.
+
+## Query
+
+```bash
+opty query "natural language description of what you're looking for"
+```
+
+Output is TOON format — compact, LLM-optimized:
+```
+functions[N]{name,signature,file,line}:
+functionName,signature,path/to/file.ts,42
+...
+```
+
+## Other Commands
+
+```bash
+opty status       # Show indexed file/unit count and watched directory
+opty reindex      # Force re-scan after major file changes or stale results
+```
+
+## Tips
+
+- Query by concept, not exact names: `"connection pool exhaustion"` not `"ConnectionPoolError"`
+- Results include file + line number — use `read` to get the actual code
+- `opty status` shows which directory is indexed (the daemon is per-project)
+- After large refactors, run `opty reindex` if results seem stale

pi/.pi/agent/skills/qmd/SKILL.md

@@ -0,0 +1,89 @@
+---
+name: qmd
+description: Search markdown knowledge bases, notes, and documentation using the qmd CLI. Use when searching notes, finding documents, or looking up information in the knowledge base.
+license: MIT
+compatibility: Requires qmd CLI. Install via `npm install -g @tobilu/qmd`.
+metadata:
+  author: tobi
+  version: "2.0.0"
+allowed-tools: Bash(qmd:*)
+---
+
+# QMD - Quick Markdown Search
+
+Local search engine for markdown content. Use via the `qmd` CLI (bash).
+
+## Status
+
+!`qmd status 2>/dev/null || echo "Not installed: npm install -g @tobilu/qmd"`
+
+## CLI Usage
+
+```bash
+qmd query "question"                        # Auto-expand + rerank (recommended)
+qmd query $'lex: X\nvec: Y'                 # Structured query document
+qmd query $'expand: question'              # Explicit expand
+qmd search "keywords"                       # BM25 only (no LLM)
+qmd get "path/to/file.md"                   # Retrieve a document by path
+qmd get "#abc123"                           # Retrieve by docid
+qmd get "path/to/file.md:100" -l 50         # Retrieve starting at line 100, 50 lines
+qmd multi-get "journals/2026-*.md" -l 40    # Batch fetch by glob
+qmd multi-get notes/foo.md,notes/bar.md     # Comma-separated list
+```
+
+## Query Types
+
+| Type | Method | Input |
+|------|--------|-------|
+| `lex` | BM25 | Keywords — exact terms, names, code |
+| `vec` | Vector | Question — natural language |
+| `hyde` | Vector | Answer — hypothetical result (50-100 words) |
+
+## Writing Good Queries
+
+**lex (keyword)**
+- 2-5 terms, no filler words
+- Exact phrase: `"connection pool"` (quoted)
+- Exclude terms: `performance -sports` (minus prefix)
+- Code identifiers work: `handleError async`
+
+**vec (semantic)**
+- Full natural language question
+- Be specific: `"how does the rate limiter handle burst traffic"`
+
+**hyde (hypothetical document)**
+- Write 50-100 words of what the *answer* looks like
+- Use the vocabulary you expect in the result
+
+**expand (auto-expand)**
+- Single-line query (implicit) or `expand: question`
+- Lets a local LLM generate lex/vec/hyde variations
+- Do not mix with other typed lines
+
+## Combining Types
+
+| Goal | Approach |
+|------|----------|
+| Know exact terms | `lex` only |
+| Don't know vocabulary | Single-line (implicit expand) or `vec` |
+| Best recall | `lex` + `vec` |
+| Complex topic | `lex` + `vec` + `hyde` |
+
+First query gets 2x weight — put your strongest signal first.
+
+## Lex Syntax
+
+| Syntax | Meaning | Example |
+|--------|---------|---------|
+| `term` | Prefix match | `perf` matches "performance" |
+| `"phrase"` | Exact phrase | `"rate limiter"` |
+| `-term` | Exclude | `performance -sports` |
+
+## Collection Filtering
+
+```bash
+qmd query --collections docs "question"
+qmd query --collections docs,notes "question"
+```
+
+Omit to search all collections.

pi/.pi/web-search.json

@@ -1,3 +1,3 @@
 {
-  "geminiApiKey": "AIzaSyAwY4Fyq9-9EvtjOqUAWP2TIYc02QxR9w8"
+  "provider": "auto"
 }