@probelabs/visor 0.1.181-ee → 0.1.182-ee

package/defaults/code-talk.yaml

@@ -136,18 +136,59 @@ outputs:
     value_js: |
       const result = outputs?.['explore-code'];
       if (result?.answer) return result.answer;
-      const routeOutput = outputs?.['setup-projects']?.routing_decision;
-      // Handle proper notes field
-      const routeNotes = routeOutput?.notes;
-      if (typeof routeNotes === 'string' && routeNotes.trim().length > 0) {
-        return { text: routeNotes };
+      const resultText = typeof result?.text === 'string' ? result.text.trim() : '';
+      if (resultText.length > 0) {
+        return { text: resultText };
       }
-      // Fallback: if AI returned {text: "..."} instead of proper schema
-      const routeText = routeOutput?.text;
-      if (typeof routeText === 'string' && routeText.trim().length > 0) {
-        return { text: routeText };
+      const routeNotes = outputs?.['setup-projects']?.routing_decision?.notes;
+      const checkoutProjects = outputs?.['setup-projects']?.checkout_projects;
+      if (
+        (!Array.isArray(checkoutProjects) || checkoutProjects.length === 0) &&
+        typeof routeNotes === 'string' &&
+        routeNotes.trim().length > 0
+      ) {
+        return { text: routeNotes.trim() };
+      }
+      return { text: 'Code exploration did not produce an answer.' };
+
+  - name: exploration_status
+    description: Outcome of the exploration step
+    value_js: |
+      const result = outputs?.['explore-code'];
+      if (result?.answer?.text) return 'success';
+      const resultText = typeof result?.text === 'string' ? result.text.trim() : '';
+      if (resultText.length > 0) {
+        if (/timed out/i.test(resultText)) return 'timeout';
+        return 'failed';
       }
-      return null;
+      const routeNotes = outputs?.['setup-projects']?.routing_decision?.notes;
+      const checkoutProjects = outputs?.['setup-projects']?.checkout_projects;
+      if (
+        (!Array.isArray(checkoutProjects) || checkoutProjects.length === 0) &&
+        typeof routeNotes === 'string' &&
+        routeNotes.trim().length > 0
+      ) {
+        return 'no_projects';
+      }
+      return 'failed';
+
+  - name: exploration_error
+    description: Timeout or failure detail when exploration did not return a real answer
+    value_js: |
+      const result = outputs?.['explore-code'];
+      if (result?.answer?.text) return '';
+      const resultText = typeof result?.text === 'string' ? result.text.trim() : '';
+      if (resultText.length > 0) return resultText;
+      const routeNotes = outputs?.['setup-projects']?.routing_decision?.notes;
+      const checkoutProjects = outputs?.['setup-projects']?.checkout_projects;
+      if (
+        (!Array.isArray(checkoutProjects) || checkoutProjects.length === 0) &&
+        typeof routeNotes === 'string' &&
+        routeNotes.trim().length > 0
+      ) {
+        return routeNotes.trim();
+      }
+      return 'Code exploration did not produce an answer.';
 
   - name: references
     description: Code/doc references from exploration
@@ -174,11 +215,20 @@ outputs:
       const result = outputs?.['explore-code'];
       const confidence = result?.confidence;
       const reason = result?.confidence_reason;
-      if (typeof reason === 'string') return reason;
+      if (typeof reason === 'string' && reason.trim().length > 0) return reason;
       if (confidence === 'high') return '';
+      const resultText = typeof result?.text === 'string' ? result.text.trim() : '';
+      if (resultText.length > 0) return resultText;
       const routeNotes = outputs?.['setup-projects']?.routing_decision?.notes;
-      if (typeof routeNotes === 'string' && routeNotes.trim().length > 0) return routeNotes;
-      return 'No confidence explanation was provided by explore-code.';
+      const checkoutProjects = outputs?.['setup-projects']?.checkout_projects;
+      if (
+        (!Array.isArray(checkoutProjects) || checkoutProjects.length === 0) &&
+        typeof routeNotes === 'string' &&
+        routeNotes.trim().length > 0
+      ) {
+        return routeNotes.trim();
+      }
+      return 'Code exploration did not produce an answer.';
 
   - name: projects_explored
     description: Which project IDs were checked out
@@ -261,7 +311,7 @@ steps:
       skip_code_context: true
       enableDelegate: true
       enableExecutePlan: false
-      max_iterations: 50
+      max_iterations: 100
       prompt_type: code-explorer
       allowBash: true
       bashConfig:
@@ -415,8 +465,17 @@ steps:
       - Each delegate should answer ONE specific question (not "look at the code")
       - Run multiple delegates in PARALLEL for different hypotheses or components
       - Ask delegates to return specific file paths and line numbers
+      - Do NOT delegate or re-search the same question twice in one investigation
+      - If a delegate returns enough evidence for the current claim, stop and use it
 
       Relay complete data from tools — do not summarize or compress tool output.
+
+      Investigation scope:
+      - Stop once you have enough evidence to answer the question accurately
+      - If this is an implementation handoff for engineer, optimize for the minimum
+        sufficient handoff: repo, branch/ref, target files, relevant tests, and the
+        key evidence explaining why those files matter
+      - Prefer one search followed by targeted extract over repeated broad searches
       </instructions>
 
       {% if inputs.exploration_prompt %}
@@ -468,6 +527,13 @@ steps:
       implementation, then consult docs to confirm semantics. When multiple projects
       are involved, trace data and config flow across them.
 
+      Efficiency rules for this investigation:
+      - Reuse evidence already found in earlier tool results
+      - If the question is narrow and the relevant files are already identified,
+        stop exploring and answer
+      - If the next consumer is engineer, avoid broad code archaeology once the
+        implementation target and validation path are clear
+
       Synthesize a single answer:
       - Ground everything in code/docs evidence
       - End with a "## References" section with clickable GitHub links:

package/defaults/engineer.yaml

@@ -338,19 +338,25 @@ steps:
       <delegation>
       Use the delegate tool for parallel work, plan validation, and build discovery.
 
-      MANDATORY FIRST STEP — delegate "Discover build system" for each repository:
+      FIRST decide whether delegation is needed.
+
+      Delegate "Discover build system" ONLY when the exact commands are not already
+      available in the provided context, code-explorer output, project metadata, or
+      recent tool results.
       - Check: Makefile, package.json (scripts), Cargo.toml, go.mod, pyproject.toml
       - Check: CI config (.github/workflows/, .gitlab-ci.yml, Jenkinsfile)
       - Check: README for build/test/lint instructions
       - Return the EXACT commands for: build, test, lint/format, and any pre-commit hooks
       - Example output: "build: make, test: make test, lint: gofmt -l . && golangci-lint run"
-      Use these commands throughout the session. Do NOT guess — use what the delegate found.
+      - Reuse these commands throughout the session. Do NOT rediscover them once known.
 
-      MANDATORY BEFORE IMPLEMENTATION — delegate "Plan validation":
+      Delegate "Plan validation" ONLY when the task is broad, high-risk, multi-repo,
+      or the implementation path is still unclear after reviewing existing context.
       - Describe: files to change, approach, patterns to follow
       - Ask the delegate to verify: do these files exist? Are there existing tests?
         Are there related utilities or patterns to reuse? Any API contracts to respect?
-      - Wait for the response before writing code
+      - Skip this delegate for narrow single-repo changes when code-explorer or direct
+        inspection already identified the target files, branch, and validation path.
 
       Also delegate for:
       - Multi-repo changes (one delegate per repo, in parallel)
@@ -361,9 +367,12 @@ steps:
       - Sequential dependent work (step B needs step A's output)
       - Simple single-file edits (fewer than 5 iterations)
       - Git operations (commit, push, PR) — always do these yourself
+      - Questions you already delegated once in this session
 
       Delegates have fewer iterations and no access to your conversation.
       Provide all necessary context in the delegate prompt.
+      If a delegate returns empty output, times out, or repeats information already
+      known, do NOT call the same delegate again. Fall back to direct tools.
       </delegation>
 
       <git-workflow>
@@ -372,12 +381,17 @@ steps:
 
       Before your final response, verify:
       □ Build passes (using exact commands from "Discover build system" delegate)
-      □ Tests pass (run the full test suite, not just your new tests)
+      □ Tests pass (start with the narrowest relevant tests; run broader suites only
+        when required by repo policy, when the change is cross-cutting, or when focused
+        tests indicate wider impact)
       □ Lint/format passes (if the project has a linter)
       □ git add <files>
       □ git commit -m "descriptive message"
       □ git push -u origin <branch-name>
       □ gh pr create (for new PRs) or update existing PR
+        - For a new branch: ALWAYS push first, then use `gh pr create --head <branch-name>`
+        - If PR creation fails, inspect stderr, fix the missing prerequisite, and retry once
+        - Do NOT repeat the same `gh pr create` command after the same error
       No PR URL = failed task. Report errors honestly, never claim false success.
 
       If build/test/lint fails, fix the issue before committing. If you cannot fix it,
@@ -397,19 +411,23 @@ steps:
       <efficiency>
       - Use context data directly — don't re-read files or re-run searches for
         information already provided by code-explorer.
+      - If code-explorer already identified the repo, branch, files, tests, or exact
+        commands, treat that as the default source of truth unless a tool result proves
+        it wrong.
       - If a project has <setup> commands listed, run them FIRST (in the project's
         directory) before any other work. These are prerequisites (e.g., `npm install`,
         `make deps`, database migrations).
-      - Use tasks to track multi-step work. Create these tasks at minimum:
-        1. "Run setup commands" — execute <setup> commands for each project (if any)
-        2. "Discover build system" (delegate) — find exact build/test/lint commands
-        3. "Plan validation" (delegate) — verify approach before coding
-        4. "Implement changes" — the actual code/file changes
-        5. "Verify build" — run build, test, and lint commands; fix any failures
-        6. "Create pull request" — git branch, add, commit, push, gh pr create
-        Mark in_progress/completed as you go. Do NOT skip "Verify build".
-      - If a bash command fails, try a different approach. Don't retry the same
-        command or get stuck in loops.
+      - Use tasks to track real phases of work, not every obvious micro-step.
+      - For narrow single-repo changes, keep the task list minimal:
+        1. "Implement changes"
+        2. "Verify build"
+        3. "Create pull request"
+      - Add "Run setup commands", "Discover build system", or "Plan validation"
+        ONLY when you actually need to perform those steps.
+      - Mark in_progress/completed as you go. Do NOT skip "Verify build".
+      - If a bash command fails, diagnose the cause before retrying.
+      - Do NOT repeat the same logical action after the same error unless you changed
+        a prerequisite (for example: push before re-running `gh pr create`).
       </efficiency>
       {% assign has_trace = inputs.trace_id | size %}
       {% assign has_slack_user = inputs.slack_user_id | size %}

package/defaults/skills/code-explorer.yaml

@@ -29,6 +29,11 @@ knowledge: |
   - If confidence "high", trust the answer — do NOT re-call with rephrased question
   - Only call again for a genuinely DIFFERENT aspect of the codebase
   - If confidence "medium" or "low", check confidence_reason for what to refine
+  - If `exploration_status` is `timeout`, `failed`, or `no_projects`, do NOT re-call
+    with a paraphrase of the same question. Report the failure honestly and only retry
+    if you can narrow the question or change the scope.
+  - If `references` is empty and confidence is low, treat that as "not answered yet",
+    not as a usable code answer.
 
   ## Usage Instructions
   1. Call the `code-explorer` tool with the user's question — do NOT try to answer code questions yourself

package/dist/agent-protocol/a2a-frontend.d.ts

@@ -39,6 +39,16 @@ export declare class A2AFrontend implements ActiveFrontend {
     /** Set the full Visor config (needed for check definitions). */
     setVisorConfig(config: VisorConfig): void;
     start(ctx: FrontendContext): Promise<void>;
+    /**
+     * Stop listening for new connections and free the port.
+     * In-flight tasks continue processing.
+     */
+    stopListening(): Promise<void>;
+    /**
+     * Drain: stop accepting new tasks, wait for in-flight tasks to complete.
+     * @param timeoutMs - Max wait time. 0 = unlimited (default).
+     */
+    drain(timeoutMs?: number): Promise<void>;
     stop(): Promise<void>;
     private handleRequest;
     private serveAgentCard;

package/dist/agent-protocol/a2a-frontend.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"a2a-frontend.d.ts","sourceRoot":"","sources":["file:///home/runner/work/visor/visor/src/agent-protocol/a2a-frontend.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;GAWG;AAOH,OAAO,KAAK,EAAE,cAAc,EAAE,eAAe,EAAE,MAAM,mBAAmB,CAAC;AACzE,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,cAAc,CAAC;AAE9C,OAAO,KAAK,EACV,SAAS,EACT,mBAAmB,EACnB,YAAY,EAGZ,SAAS,EACT,aAAa,EAGb,SAAS,EAGV,MAAM,SAAS,CAAC;AAajB,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,iBAAiB,CAAC;AAkGnD,wBAAgB,sBAAsB,CACpC,OAAO,EAAE,YAAY,EACrB,IAAI,EAAE,SAAS,GACd,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CA4BzB;AAED,+EAA+E;AAC/E,wBAAgB,iBAAiB,CAAC,YAAY,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAAG,aAAa,EAAE,CAmCxF;AAMD,qBAAa,WAAY,YAAW,cAAc;IAChD,QAAQ,CAAC,IAAI,SAAS;IAEtB,OAAO,CAAC,MAAM,CAA2C;IACzD,OAAO,CAAC,SAAS,CAAY;IAC7B,OAAO,CAAC,SAAS,CAA0B;IAC3C,OAAO,CAAC,MAAM,CAAsB;IACpC,OAAO,CAAC,YAAY,CAA+C;IACnE,OAAO,CAAC,IAAI,CAAgC;IAC5C,OAAO,CAAC,aAAa,CAA2B;IAChD,OAAO,CAAC,WAAW,CAAiC;IACpD,OAAO,CAAC,OAAO,CAAa;IAC5B,OAAO,CAAC,YAAY,CAA4B;IAChD,OAAO,CAAC,SAAS,CAA0B;IAE3C,OAAO,CAAC,UAAU,CAAK;gBAEX,MAAM,EAAE,mBAAmB,EAAE,SAAS,CAAC,EAAE,SAAS;IAK9D,iFAAiF;IACjF,IAAI,SAAS,IAAI,MAAM,CAEtB;IAED,sDAAsD;IACtD,SAAS,CAAC,MAAM,EAAE,GAAG,GAAG,IAAI;IAI5B,gEAAgE;IAChE,cAAc,CAAC,MAAM,EAAE,WAAW,GAAG,IAAI;IAInC,KAAK,CAAC,GAAG,EAAE,eAAe,GAAG,OAAO,CAAC,IAAI,CAAC;IAgJ1C,IAAI,IAAI,OAAO,CAAC,IAAI,CAAC;YAoBb,aAAa;IA2G3B,OAAO,CAAC,cAAc;YAOR,iBAAiB;YAyEjB,qBAAqB;IAwEnC,OAAO,CAAC,aAAa;IAMrB,OAAO,CAAC,eAAe;IAWvB,OAAO,CAAC,gBAAgB;YAaV,0BAA0B;IA2CxC,OAAO,CAAC,qBAAqB;YAiCf,sBAAsB;IAwBpC,OAAO,CAAC,qBAAqB;IAQ7B,OAAO,CAAC,mBAAmB;IAQ3B,OAAO,CAAC,sBAAsB;IAY9B,kFAAkF;IAClF,eAAe,CACb,MAAM,EAAE,MAAM,EACd,SAAS,EAAE,MAAM,EACjB,MAAM,EAAE;QAAE,KAAK,EAAE,SAAS,CAAC;QAAC,OAAO,CAAC,EAAE,YAAY,CAAC;QAAC,SAAS,EAAE,MAAM,CAAA;KAAE,GACtE,IAAI;IAaP,gFAAgF;IAChF,iBAAiB,CACf,MAAM,EAAE,MAAM,EACd,SAAS,EAAE,MAAM,EACjB,QAAQ,EAAE,aAAa,EACvB,MAAM,EAAE,OAAO,EACf,SAAS,EAAE,OAAO,GACjB,IAAI;IAmBP;;OAEG;YACW,mBAAmB;IA+CjC;;;OAGG;YACW,oBAAoB;IA8HlC;;;;OAIG;IACH,OAAO,CAAC,kBAAkB;IA2B1B,OAAO,CAAC,qBAAqB;IAoB7B,OAAO,CAAC,iBAAiB;IAazB,OAAO,CAAC,iBAAiB;IAoBzB,OAAO,CAAC,gBAAgB;IAWxB,YAAY,IAAI,SAAS;IAIzB,YAAY,IAAI,SAAS,GAAG,IAAI;CAGjC"}
+{"version":3,"file":"a2a-frontend.d.ts","sourceRoot":"","sources":["file:///home/runner/work/visor/visor/src/agent-protocol/a2a-frontend.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;GAWG;AAOH,OAAO,KAAK,EAAE,cAAc,EAAE,eAAe,EAAE,MAAM,mBAAmB,CAAC;AACzE,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,cAAc,CAAC;AAE9C,OAAO,KAAK,EACV,SAAS,EACT,mBAAmB,EACnB,YAAY,EAGZ,SAAS,EACT,aAAa,EAGb,SAAS,EAGV,MAAM,SAAS,CAAC;AAajB,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,iBAAiB,CAAC;AAkGnD,wBAAgB,sBAAsB,CACpC,OAAO,EAAE,YAAY,EACrB,IAAI,EAAE,SAAS,GACd,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CA4BzB;AAED,+EAA+E;AAC/E,wBAAgB,iBAAiB,CAAC,YAAY,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAAG,aAAa,EAAE,CAmCxF;AAMD,qBAAa,WAAY,YAAW,cAAc;IAChD,QAAQ,CAAC,IAAI,SAAS;IAEtB,OAAO,CAAC,MAAM,CAA2C;IACzD,OAAO,CAAC,SAAS,CAAY;IAC7B,OAAO,CAAC,SAAS,CAA0B;IAC3C,OAAO,CAAC,MAAM,CAAsB;IACpC,OAAO,CAAC,YAAY,CAA+C;IACnE,OAAO,CAAC,IAAI,CAAgC;IAC5C,OAAO,CAAC,aAAa,CAA2B;IAChD,OAAO,CAAC,WAAW,CAAiC;IACpD,OAAO,CAAC,OAAO,CAAa;IAC5B,OAAO,CAAC,YAAY,CAA4B;IAChD,OAAO,CAAC,SAAS,CAA0B;IAE3C,OAAO,CAAC,UAAU,CAAK;gBAEX,MAAM,EAAE,mBAAmB,EAAE,SAAS,CAAC,EAAE,SAAS;IAK9D,iFAAiF;IACjF,IAAI,SAAS,IAAI,MAAM,CAEtB;IAED,sDAAsD;IACtD,SAAS,CAAC,MAAM,EAAE,GAAG,GAAG,IAAI;IAI5B,gEAAgE;IAChE,cAAc,CAAC,MAAM,EAAE,WAAW,GAAG,IAAI;IAInC,KAAK,CAAC,GAAG,EAAE,eAAe,GAAG,OAAO,CAAC,IAAI,CAAC;IAgJhD;;;OAGG;IACG,aAAa,IAAI,OAAO,CAAC,IAAI,CAAC;IAWpC;;;OAGG;IACG,KAAK,CAAC,SAAS,SAAI,GAAG,OAAO,CAAC,IAAI,CAAC;IAoBnC,IAAI,IAAI,OAAO,CAAC,IAAI,CAAC;YAoBb,aAAa;IA2G3B,OAAO,CAAC,cAAc;YAOR,iBAAiB;YAyEjB,qBAAqB;IAwEnC,OAAO,CAAC,aAAa;IAMrB,OAAO,CAAC,eAAe;IAWvB,OAAO,CAAC,gBAAgB;YAaV,0BAA0B;IA2CxC,OAAO,CAAC,qBAAqB;YAiCf,sBAAsB;IAwBpC,OAAO,CAAC,qBAAqB;IAQ7B,OAAO,CAAC,mBAAmB;IAQ3B,OAAO,CAAC,sBAAsB;IAY9B,kFAAkF;IAClF,eAAe,CACb,MAAM,EAAE,MAAM,EACd,SAAS,EAAE,MAAM,EACjB,MAAM,EAAE;QAAE,KAAK,EAAE,SAAS,CAAC;QAAC,OAAO,CAAC,EAAE,YAAY,CAAC;QAAC,SAAS,EAAE,MAAM,CAAA;KAAE,GACtE,IAAI;IAaP,gFAAgF;IAChF,iBAAiB,CACf,MAAM,EAAE,MAAM,EACd,SAAS,EAAE,MAAM,EACjB,QAAQ,EAAE,aAAa,EACvB,MAAM,EAAE,OAAO,EACf,SAAS,EAAE,OAAO,GACjB,IAAI;IAmBP;;OAEG;YACW,mBAAmB;IA+CjC;;;OAGG;YACW,oBAAoB;IA8HlC;;;;OAIG;IACH,OAAO,CAAC,kBAAkB;IA2B1B,OAAO,CAAC,qBAAqB;IAoB7B,OAAO,CAAC,iBAAiB;IAazB,OAAO,CAAC,iBAAiB;IAoBzB,OAAO,CAAC,gBAAgB;IAWxB,YAAY,IAAI,SAAS;IAIzB,YAAY,IAAI,SAAS,GAAG,IAAI;CAGjC"}

package/dist/agent-protocol/task-evaluator.d.ts

@@ -0,0 +1,52 @@
+/**
+ * Task Response Evaluator with LLM Judge + Trace Analysis.
+ *
+ * Evaluates completed agent tasks on two axes:
+ * 1. Response quality (relevance, completeness, actionability)
+ * 2. Execution quality (tool call efficiency, unnecessary delegations)
+ *
+ * Uses ProbeAgent in single-shot mode (same pattern as llm-judge.ts).
+ */
+import type { SqliteTaskStore } from './task-store';
+export interface ResponseQuality {
+    rating: number;
+    category: 'excellent' | 'good' | 'adequate' | 'poor' | 'off-topic' | 'error';
+    relevance: boolean;
+    completeness: boolean;
+    actionable: boolean;
+    reasoning: string;
+}
+export interface ExecutionQuality {
+    rating: number;
+    category: 'efficient' | 'adequate' | 'wasteful' | 'error';
+    unnecessary_tool_calls?: number;
+    reasoning: string;
+}
+export interface TaskEvaluationResult {
+    response_quality: ResponseQuality;
+    execution_quality?: ExecutionQuality;
+    overall_rating: number;
+    summary: string;
+}
+export interface TaskEvaluatorConfig {
+    model?: string;
+    provider?: string;
+    apiKey?: string;
+    prompt?: string;
+    traceDir?: string;
+}
+export declare const DEFAULT_EVALUATION_PROMPT = "You are a task response quality evaluator for an AI agent system called Visor.\n\nYou will receive the user's original request and an execution trace inside <execution_trace> tags. The trace is a YAML-formatted view of the entire agent execution, including the final response. When no trace is available, the agent response is provided directly.\n\n## How to Read the Execution Trace\n\nThe trace is a tree of spans representing the agent's execution pipeline:\n\n**Top-level: `visor.run`** \u2014 The root span with metadata:\n- `trace_id`: Unique execution identifier\n- `visor` / `probe`: Software versions\n- `source`: Where the request came from (e.g., \"slack\", \"cli\")\n- `duration`: Total wall-clock time\n\n**Checks** \u2014 Named processing steps (e.g., `route-intent`, `explore-code`, `generate-response`):\n- `type`: \"ai\" (LLM-powered), \"script\" (deterministic), or \"workflow\" (sub-pipeline)\n- `duration`: How long this step took\n- `input`: What was passed to this check \u2014 may include an `intent` (the user's question as understood by the router) and dependency outputs\n- `output`: The check's result \u2014 may be structured JSON or plain text\n\n**AI blocks** (`ai: model-name`) \u2014 Individual LLM calls within checks:\n- Shows model used, duration, and token counts (input/output)\n- `intent`: The question or instruction sent to the LLM\n\n**Tool calls** \u2014 Listed as `- toolName(input) \u2192 size`:\n- `search(\"query\" in repo)`: Code search. \"\u2192 no results\" means nothing was found; otherwise shows result size\n- `extract(file/path)`: File content extraction with result size\n- `listFiles(dir)`: Directory listing\n- `bash()`: Shell command execution\n\n**Delegations** (`search.delegate(\"query\")`) \u2014 Sub-agent searches:\n- Contains their own AI blocks and tool calls\n- Used for complex multi-step code exploration\n\n**The `response` field** at the end of the trace is the final answer sent back to the user. This is the primary output to evaluate.\n\n**Symbols:**\n- `\u2717` marks failed/error spans\n- `= check-name` means output is identical to that check's output (deduplication)\n\n## Evaluation Criteria\n\n**Response Quality** (1-5):\n- **Relevance**: Does the response directly address what the user asked? A response about the wrong topic or that misunderstands the question scores low.\n- **Completeness**: Does it fully answer the question? Partial answers, missing key details, or surface-level responses score lower.\n- **Actionable**: Can the user act on this information? Vague or generic advice scores lower than specific, concrete answers with code references.\n- Rating: 5=excellent (thorough, specific, directly useful), 4=good (answers well but minor gaps), 3=adequate (addresses question but lacks depth), 2=poor (partially relevant or very incomplete), 1=off-topic or error\n\n**Execution Quality** (1-5, only when trace is provided):\n- **Efficiency**: Were tool calls necessary and well-targeted? Good search queries that find results on the first try score high.\n- **Redundancy**: Were there duplicate searches, unnecessary re-searches with slightly different queries, or tools called for information already available?\n- **Delegation quality**: Were search delegations productive? Did they explore relevant code paths?\n- **Token usage**: Was input context kept reasonable, or did the agent load excessive amounts of code?\n- Rating: 5=efficient (minimal, targeted tool use), 4=adequate (minor redundancy), 3=some waste (noticeable unnecessary calls), 2=wasteful (many redundant searches or delegations), 1=error/broken execution\n\n**Overall Rating** (1-5): Weighted combination \u2014 response quality matters most, execution quality is secondary. A perfect response from a wasteful execution still scores 3-4 overall.\n\nYou MUST respond with valid JSON matching the provided schema. Be specific in your reasoning \u2014 reference actual check names, tool calls, or response content.";
+/**
+ * Evaluate a completed task using an LLM judge.
+ *
+ * @param taskId - Task ID (full or prefix)
+ * @param store - Initialized SqliteTaskStore
+ * @param config - Optional evaluator configuration
+ * @returns Evaluation result with ratings and reasoning
+ */
+export declare function evaluateTask(taskId: string, store: SqliteTaskStore, config?: TaskEvaluatorConfig): Promise<TaskEvaluationResult>;
+/**
+ * Run evaluation and store result as a task artifact.
+ */
+export declare function evaluateAndStore(taskId: string, store: SqliteTaskStore, config?: TaskEvaluatorConfig): Promise<TaskEvaluationResult>;
+//# sourceMappingURL=task-evaluator.d.ts.map

package/dist/agent-protocol/task-evaluator.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"task-evaluator.d.ts","sourceRoot":"","sources":["file:///home/runner/work/visor/visor/src/agent-protocol/task-evaluator.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AAIH,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,cAAc,CAAC;AAOpD,MAAM,WAAW,eAAe;IAC9B,MAAM,EAAE,MAAM,CAAC;IACf,QAAQ,EAAE,WAAW,GAAG,MAAM,GAAG,UAAU,GAAG,MAAM,GAAG,WAAW,GAAG,OAAO,CAAC;IAC7E,SAAS,EAAE,OAAO,CAAC;IACnB,YAAY,EAAE,OAAO,CAAC;IACtB,UAAU,EAAE,OAAO,CAAC;IACpB,SAAS,EAAE,MAAM,CAAC;CACnB;AAED,MAAM,WAAW,gBAAgB;IAC/B,MAAM,EAAE,MAAM,CAAC;IACf,QAAQ,EAAE,WAAW,GAAG,UAAU,GAAG,UAAU,GAAG,OAAO,CAAC;IAC1D,sBAAsB,CAAC,EAAE,MAAM,CAAC;IAChC,SAAS,EAAE,MAAM,CAAC;CACnB;AAED,MAAM,WAAW,oBAAoB;IACnC,gBAAgB,EAAE,eAAe,CAAC;IAClC,iBAAiB,CAAC,EAAE,gBAAgB,CAAC;IACrC,cAAc,EAAE,MAAM,CAAC;IACvB,OAAO,EAAE,MAAM,CAAC;CACjB;AAED,MAAM,WAAW,mBAAmB;IAClC,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,QAAQ,CAAC,EAAE,MAAM,CAAC;CACnB;AAMD,eAAO,MAAM,yBAAyB,44HAyDwH,CAAC;AAoD/J;;;;;;;GAOG;AACH,wBAAsB,YAAY,CAChC,MAAM,EAAE,MAAM,EACd,KAAK,EAAE,eAAe,EACtB,MAAM,CAAC,EAAE,mBAAmB,GAC3B,OAAO,CAAC,oBAAoB,CAAC,CAqI/B;AAMD;;GAEG;AACH,wBAAsB,gBAAgB,CACpC,MAAM,EAAE,MAAM,EACd,KAAK,EAAE,eAAe,EACtB,MAAM,CAAC,EAAE,mBAAmB,GAC3B,OAAO,CAAC,oBAAoB,CAAC,CAe/B"}

package/dist/agent-protocol/task-store.d.ts

@@ -60,8 +60,10 @@ export interface TaskStore {
         rows: TaskQueueRow[];
         total: number;
     };
-    /** Mark all 'working' tasks as 'failed' (crash recovery on startup). */
-    failStaleTasks(reason?: string): number;
+    /** Mark 'working' tasks as 'failed' (crash recovery on startup).
+     *  When `claimedBy` is provided, only fails tasks owned by that instance.
+     *  Without it, only fails **unclaimed** working tasks (safe default). */
+    failStaleTasks(reason?: string, claimedBy?: string): number;
     /** Mark 'working' tasks older than the given age as 'failed' (runtime stale detection). */
     failStaleTasksByAge(olderThanMs: number, reason?: string): number;
     /** Delete completed/failed/canceled tasks older than the given age. */
@@ -95,7 +97,7 @@ export declare class SqliteTaskStore implements TaskStore {
     claimNextSubmitted(workerId: string): AgentTask | null;
     reclaimStaleTasks(_workerId: string, staleTimeoutMs?: number): AgentTask[];
     releaseClaim(taskId: string): void;
-    failStaleTasks(reason?: string): number;
+    failStaleTasks(reason?: string, claimedBy?: string): number;
     failStaleTasksByAge(olderThanMs: number, reason?: string): number;
     purgeOldTasks(olderThanMs: number): number;
     deleteExpiredTasks(): string[];

package/dist/agent-protocol/task-store.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"task-store.d.ts","sourceRoot":"","sources":["file:///home/runner/work/visor/visor/src/agent-protocol/task-store.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AAMH,OAAO,EACL,SAAS,EACT,YAAY,EACZ,aAAa,EACb,sBAAsB,EACtB,SAAS,EAGV,MAAM,SAAS,CAAC;AAgFjB,MAAM,WAAW,gBAAgB;IAC/B,SAAS,EAAE,MAAM,CAAC;IAClB,cAAc,EAAE,YAAY,CAAC;IAC7B,aAAa,CAAC,EAAE,sBAAsB,CAAC;IACvC,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IAC1C,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,eAAe;IAC9B,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,KAAK,CAAC,EAAE,SAAS,EAAE,CAAC;IACpB,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,MAAM,CAAC,EAAE,MAAM,CAAC;CACjB;AAED,mEAAmE;AACnE,MAAM,WAAW,YAAY;IAC3B,EAAE,EAAE,MAAM,CAAC;IACX,UAAU,EAAE,MAAM,CAAC;IACnB,KAAK,EAAE,SAAS,CAAC;IACjB,UAAU,EAAE,MAAM,CAAC;IACnB,UAAU,EAAE,MAAM,CAAC;IACnB,UAAU,EAAE,MAAM,GAAG,IAAI,CAAC;IAC1B,UAAU,EAAE,MAAM,GAAG,IAAI,CAAC;IAC1B,WAAW,EAAE,MAAM,GAAG,IAAI,CAAC;IAC3B,MAAM,EAAE,MAAM,GAAG,IAAI,CAAC;IACtB,eAAe,EAAE,MAAM,CAAC;IACxB,MAAM,EAAE,MAAM,CAAC;IACf,QAAQ,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CACnC;AAED,MAAM,WAAW,eAAe;IAC9B,KAAK,EAAE,SAAS,EAAE,CAAC;IACnB,KAAK,EAAE,MAAM,CAAC;CACf;AAED,MAAM,WAAW,SAAS;IACxB,UAAU,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC;IAC5B,QAAQ,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC;IAG1B,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,SAAS,CAAC;IAChD,OAAO,CAAC,MAAM,EAAE,MAAM,GAAG,SAAS,GAAG,IAAI,CAAC;IAC1C,SAAS,CAAC,MAAM,EAAE,eAAe,GAAG,eAAe,CAAC;IAGpD,eAAe,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,SAAS,EAAE,aAAa,CAAC,EAAE,YAAY,GAAG,IAAI,CAAC;IACzF,0FAA0F;IAC1F,SAAS,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM,GAAG,IAAI,CAAC;IAClD,WAAW,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,aAAa,GAAG,IAAI,CAAC;IAC3D,aAAa,CAAC,MAAM,EAAE,MAAM,EAAE,OAAO,EAAE,YAAY,GAAG,IAAI,CAAC;IAC3D,QAAQ,CAAC,MAAM,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,GAAG,IAAI,CAAC;IAG9C,kBAAkB,CAAC,QAAQ,EAAE,MAAM,GAAG,SAAS,GAAG,IAAI,CAAC;IACvD,iBAAiB,CAAC,QAAQ,EAAE,MAAM,EAAE,cAAc,CAAC,EAAE,MAAM,GAAG,SAAS,EAAE,CAAC;IAC1E,YAAY,CAAC,MAAM,EAAE,MAAM,GAAG,IAAI,CAAC;IAGnC,YAAY,CAAC,CAAC,MAAM,EAAE,eAAe,GAAG;QAAE,IAAI,EAAE,YAAY,EAAE,CAAC;QAAC,KAAK,EAAE,MAAM,CAAA;KAAE,CAAC;IAGhF,wEAAwE;IACxE,cAAc,CAAC,MAAM,CAAC,EAAE,MAAM,GAAG,MAAM,CAAC;IACxC,2FAA2F;IAC3F,mBAAmB,CAAC,WAAW,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,MAAM,GAAG,MAAM,CAAC;IAClE,uEAAuE;IACvE,aAAa,CAAC,WAAW,EAAE,MAAM,GAAG,MAAM,CAAC;IAC3C,kBAAkB,IAAI,MAAM,EAAE,CAAC;IAC/B,UAAU,CAAC,MAAM,EAAE,MAAM,GAAG,IAAI,CAAC;CAClC;AAMD,qBAAa,eAAgB,YAAW,SAAS;IAC/C,OAAO,CAAC,EAAE,CAAqC;IAC/C,OAAO,CAAC,MAAM,CAAS;gBAEX,QAAQ,CAAC,EAAE,MAAM;IAIvB,UAAU,IAAI,OAAO,CAAC,IAAI,CAAC;IA4B3B,QAAQ,IAAI,OAAO,CAAC,IAAI,CAAC;IAO/B,OAAO,CAAC,KAAK;IAKb,0FAA0F;IAC1F,WAAW,IAAI,OAAO;IAItB,OAAO,CAAC,aAAa;IAkCrB,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,SAAS;IAuC/C,OAAO,CAAC,MAAM,EAAE,MAAM,GAAG,SAAS,GAAG,IAAI;IAQzC,OAAO,CAAC,iBAAiB;IAgCzB,SAAS,CAAC,MAAM,EAAE,eAAe,GAAG,eAAe;IAqBnD,YAAY,CAAC,MAAM,EAAE,eAAe,GAAG;QAAE,IAAI,EAAE,YAAY,EAAE,CAAC;QAAC,KAAK,EAAE,MAAM,CAAA;KAAE;IAoD9E,eAAe,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,SAAS,EAAE,aAAa,CAAC,EAAE,YAAY,GAAG,IAAI;IAkBxF,SAAS,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM,GAAG,IAAI;IAQjD,WAAW,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,aAAa,GAAG,IAAI;IAkB1D,aAAa,CAAC,MAAM,EAAE,MAAM,EAAE,OAAO,EAAE,YAAY,GAAG,IAAI;IAkB1D,QAAQ,CAAC,MAAM,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,GAAG,IAAI;IAa7C,kBAAkB,CAAC,QAAQ,EAAE,MAAM,GAAG,SAAS,GAAG,IAAI;IAyBtD,iBAAiB,CAAC,SAAS,EAAE,MAAM,EAAE,cAAc,CAAC,EAAE,MAAM,GAAG,SAAS,EAAE;IAqB1E,YAAY,CAAC,MAAM,EAAE,MAAM,GAAG,IAAI;IAalC,cAAc,CAAC,MAAM,CAAC,EAAE,MAAM,GAAG,MAAM;IAmBvC,mBAAmB,CAAC,WAAW,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,MAAM,GAAG,MAAM;IAqBjE,aAAa,CAAC,WAAW,EAAE,MAAM,GAAG,MAAM;IAa1C,kBAAkB,IAAI,MAAM,EAAE;IAkB9B,UAAU,CAAC,MAAM,EAAE,MAAM,GAAG,IAAI;IAShC,OAAO,CAAC,gBAAgB;CAezB"}
+{"version":3,"file":"task-store.d.ts","sourceRoot":"","sources":["file:///home/runner/work/visor/visor/src/agent-protocol/task-store.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AAMH,OAAO,EACL,SAAS,EACT,YAAY,EACZ,aAAa,EACb,sBAAsB,EACtB,SAAS,EAGV,MAAM,SAAS,CAAC;AAgFjB,MAAM,WAAW,gBAAgB;IAC/B,SAAS,EAAE,MAAM,CAAC;IAClB,cAAc,EAAE,YAAY,CAAC;IAC7B,aAAa,CAAC,EAAE,sBAAsB,CAAC;IACvC,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IAC1C,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,eAAe;IAC9B,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,KAAK,CAAC,EAAE,SAAS,EAAE,CAAC;IACpB,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,MAAM,CAAC,EAAE,MAAM,CAAC;CACjB;AAED,mEAAmE;AACnE,MAAM,WAAW,YAAY;IAC3B,EAAE,EAAE,MAAM,CAAC;IACX,UAAU,EAAE,MAAM,CAAC;IACnB,KAAK,EAAE,SAAS,CAAC;IACjB,UAAU,EAAE,MAAM,CAAC;IACnB,UAAU,EAAE,MAAM,CAAC;IACnB,UAAU,EAAE,MAAM,GAAG,IAAI,CAAC;IAC1B,UAAU,EAAE,MAAM,GAAG,IAAI,CAAC;IAC1B,WAAW,EAAE,MAAM,GAAG,IAAI,CAAC;IAC3B,MAAM,EAAE,MAAM,GAAG,IAAI,CAAC;IACtB,eAAe,EAAE,MAAM,CAAC;IACxB,MAAM,EAAE,MAAM,CAAC;IACf,QAAQ,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CACnC;AAED,MAAM,WAAW,eAAe;IAC9B,KAAK,EAAE,SAAS,EAAE,CAAC;IACnB,KAAK,EAAE,MAAM,CAAC;CACf;AAED,MAAM,WAAW,SAAS;IACxB,UAAU,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC;IAC5B,QAAQ,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC;IAG1B,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,SAAS,CAAC;IAChD,OAAO,CAAC,MAAM,EAAE,MAAM,GAAG,SAAS,GAAG,IAAI,CAAC;IAC1C,SAAS,CAAC,MAAM,EAAE,eAAe,GAAG,eAAe,CAAC;IAGpD,eAAe,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,SAAS,EAAE,aAAa,CAAC,EAAE,YAAY,GAAG,IAAI,CAAC;IACzF,0FAA0F;IAC1F,SAAS,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM,GAAG,IAAI,CAAC;IAClD,WAAW,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,aAAa,GAAG,IAAI,CAAC;IAC3D,aAAa,CAAC,MAAM,EAAE,MAAM,EAAE,OAAO,EAAE,YAAY,GAAG,IAAI,CAAC;IAC3D,QAAQ,CAAC,MAAM,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,GAAG,IAAI,CAAC;IAG9C,kBAAkB,CAAC,QAAQ,EAAE,MAAM,GAAG,SAAS,GAAG,IAAI,CAAC;IACvD,iBAAiB,CAAC,QAAQ,EAAE,MAAM,EAAE,cAAc,CAAC,EAAE,MAAM,GAAG,SAAS,EAAE,CAAC;IAC1E,YAAY,CAAC,MAAM,EAAE,MAAM,GAAG,IAAI,CAAC;IAGnC,YAAY,CAAC,CAAC,MAAM,EAAE,eAAe,GAAG;QAAE,IAAI,EAAE,YAAY,EAAE,CAAC;QAAC,KAAK,EAAE,MAAM,CAAA;KAAE,CAAC;IAGhF;;6EAEyE;IACzE,cAAc,CAAC,MAAM,CAAC,EAAE,MAAM,EAAE,SAAS,CAAC,EAAE,MAAM,GAAG,MAAM,CAAC;IAC5D,2FAA2F;IAC3F,mBAAmB,CAAC,WAAW,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,MAAM,GAAG,MAAM,CAAC;IAClE,uEAAuE;IACvE,aAAa,CAAC,WAAW,EAAE,MAAM,GAAG,MAAM,CAAC;IAC3C,kBAAkB,IAAI,MAAM,EAAE,CAAC;IAC/B,UAAU,CAAC,MAAM,EAAE,MAAM,GAAG,IAAI,CAAC;CAClC;AAMD,qBAAa,eAAgB,YAAW,SAAS;IAC/C,OAAO,CAAC,EAAE,CAAqC;IAC/C,OAAO,CAAC,MAAM,CAAS;gBAEX,QAAQ,CAAC,EAAE,MAAM;IAIvB,UAAU,IAAI,OAAO,CAAC,IAAI,CAAC;IA4B3B,QAAQ,IAAI,OAAO,CAAC,IAAI,CAAC;IAO/B,OAAO,CAAC,KAAK;IAKb,0FAA0F;IAC1F,WAAW,IAAI,OAAO;IAItB,OAAO,CAAC,aAAa;IAkCrB,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,SAAS;IAuC/C,OAAO,CAAC,MAAM,EAAE,MAAM,GAAG,SAAS,GAAG,IAAI;IAQzC,OAAO,CAAC,iBAAiB;IAgCzB,SAAS,CAAC,MAAM,EAAE,eAAe,GAAG,eAAe;IAqBnD,YAAY,CAAC,MAAM,EAAE,eAAe,GAAG;QAAE,IAAI,EAAE,YAAY,EAAE,CAAC;QAAC,KAAK,EAAE,MAAM,CAAA;KAAE;IAoD9E,eAAe,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,SAAS,EAAE,aAAa,CAAC,EAAE,YAAY,GAAG,IAAI;IAkBxF,SAAS,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM,GAAG,IAAI;IAQjD,WAAW,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,aAAa,GAAG,IAAI;IAkB1D,aAAa,CAAC,MAAM,EAAE,MAAM,EAAE,OAAO,EAAE,YAAY,GAAG,IAAI;IAkB1D,QAAQ,CAAC,MAAM,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,GAAG,IAAI;IAa7C,kBAAkB,CAAC,QAAQ,EAAE,MAAM,GAAG,SAAS,GAAG,IAAI;IAyBtD,iBAAiB,CAAC,SAAS,EAAE,MAAM,EAAE,cAAc,CAAC,EAAE,MAAM,GAAG,SAAS,EAAE;IAqB1E,YAAY,CAAC,MAAM,EAAE,MAAM,GAAG,IAAI;IAalC,cAAc,CAAC,MAAM,CAAC,EAAE,MAAM,EAAE,SAAS,CAAC,EAAE,MAAM,GAAG,MAAM;IAkC3D,mBAAmB,CAAC,WAAW,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,MAAM,GAAG,MAAM;IAqBjE,aAAa,CAAC,WAAW,EAAE,MAAM,GAAG,MAAM;IAa1C,kBAAkB,IAAI,MAAM,EAAE;IAkB9B,UAAU,CAAC,MAAM,EAAE,MAAM,GAAG,IAAI;IAShC,OAAO,CAAC,gBAAgB;CAezB"}

package/dist/agent-protocol/tasks-cli-handler.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"tasks-cli-handler.d.ts","sourceRoot":"","sources":["file:///home/runner/work/visor/visor/src/agent-protocol/tasks-cli-handler.ts"],"names":[],"mappings":"AAslBA,wBAAsB,kBAAkB,CAAC,IAAI,EAAE,MAAM,EAAE,GAAG,OAAO,CAAC,IAAI,CAAC,CA0CtE"}
+{"version":3,"file":"tasks-cli-handler.d.ts","sourceRoot":"","sources":["file:///home/runner/work/visor/visor/src/agent-protocol/tasks-cli-handler.ts"],"names":[],"mappings":"AAs7BA,wBAAsB,kBAAkB,CAAC,IAAI,EAAE,MAAM,EAAE,GAAG,OAAO,CAAC,IAAI,CAAC,CAgHtE"}

package/dist/agent-protocol/tasks-tui.d.ts

@@ -0,0 +1,34 @@
+import { SqliteTaskStore, type ListTasksFilter } from './task-store';
+export declare class TasksTUI {
+    private screen;
+    private listBox;
+    private detailBox;
+    private statusBar;
+    private store;
+    private filter;
+    private instanceId;
+    private items;
+    private total;
+    private selectedIndex;
+    private scrollOffset;
+    private mode;
+    private detailScrollOffset;
+    private detailLines;
+    private refreshInterval?;
+    private destroyed;
+    constructor(store: SqliteTaskStore, filter: ListTasksFilter);
+    private setupKeyBindings;
+    private getVisibleHeight;
+    private moveSelection;
+    private ensureVisible;
+    private scrollDetail;
+    private loadTasks;
+    private enterDetail;
+    private render;
+    private renderDetail;
+    private wrapText;
+    start(): Promise<void>;
+    destroy(): void;
+}
+export declare function runTasksTUI(filter: ListTasksFilter): Promise<void>;
+//# sourceMappingURL=tasks-tui.d.ts.map

package/dist/agent-protocol/tasks-tui.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"tasks-tui.d.ts","sourceRoot":"","sources":["file:///home/runner/work/visor/visor/src/agent-protocol/tasks-tui.ts"],"names":[],"mappings":"AAOA,OAAO,EAAE,eAAe,EAAE,KAAK,eAAe,EAAqB,MAAM,cAAc,CAAC;AAmGxF,qBAAa,QAAQ;IACnB,OAAO,CAAC,MAAM,CAAyB;IACvC,OAAO,CAAC,OAAO,CAA6B;IAC5C,OAAO,CAAC,SAAS,CAA6B;IAC9C,OAAO,CAAC,SAAS,CAA6B;IAC9C,OAAO,CAAC,KAAK,CAAkB;IAC/B,OAAO,CAAC,MAAM,CAAkB;IAChC,OAAO,CAAC,UAAU,CAAS;IAE3B,OAAO,CAAC,KAAK,CAAsB;IACnC,OAAO,CAAC,KAAK,CAAK;IAClB,OAAO,CAAC,aAAa,CAAK;IAC1B,OAAO,CAAC,YAAY,CAAK;IACzB,OAAO,CAAC,IAAI,CAAoB;IAChC,OAAO,CAAC,kBAAkB,CAAK;IAC/B,OAAO,CAAC,WAAW,CAAgB;IAEnC,OAAO,CAAC,eAAe,CAAC,CAAiB;IACzC,OAAO,CAAC,SAAS,CAAS;gBAEd,KAAK,EAAE,eAAe,EAAE,MAAM,EAAE,eAAe;IA2E3D,OAAO,CAAC,gBAAgB;IAkHxB,OAAO,CAAC,gBAAgB;IAKxB,OAAO,CAAC,aAAa;IAOrB,OAAO,CAAC,aAAa;IAYrB,OAAO,CAAC,YAAY;IAMpB,OAAO,CAAC,SAAS;IAejB,OAAO,CAAC,WAAW;IAgBnB,OAAO,CAAC,MAAM;IA2Ed,OAAO,CAAC,YAAY;IA8GpB,OAAO,CAAC,QAAQ;IAoBV,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IAmB5B,OAAO,IAAI,IAAI;CAahB;AAMD,wBAAsB,WAAW,CAAC,MAAM,EAAE,eAAe,GAAG,OAAO,CAAC,IAAI,CAAC,CAUxE"}

package/dist/agent-protocol/trace-serializer.d.ts

@@ -0,0 +1,90 @@
+/**
+ * Trace serializer for LLM evaluation prompts and CLI display.
+ *
+ * Supports multiple trace backends:
+ * 1. Grafana Tempo (via Grafana datasource proxy or direct Tempo API)
+ * 2. Jaeger (via Jaeger HTTP API)
+ * 3. Local NDJSON files (Visor's file-span-exporter OTEL format)
+ *
+ * Auto-detects the backend from environment variables:
+ * - OTEL_EXPORTER_OTLP_ENDPOINT + VISOR_TELEMETRY_SINK=otlp → probe Grafana/Jaeger
+ * - VISOR_TRACE_BACKEND=grafana|jaeger|file (explicit override)
+ * - GRAFANA_URL, GRAFANA_TEMPO_DATASOURCE_ID → Grafana Tempo
+ * - JAEGER_URL → Jaeger
+ * - Falls back to local file scan
+ */
+export interface TraceBackendConfig {
+    type: 'grafana' | 'jaeger' | 'file' | 'auto';
+    /** Grafana base URL, e.g. http://localhost:3000 */
+    grafanaUrl?: string;
+    /** Grafana Tempo datasource ID (default: auto-detect) */
+    grafanaDatasourceId?: string | number;
+    /** Jaeger base URL, e.g. http://localhost:16686 */
+    jaegerUrl?: string;
+    /** Local trace directory (default: output/traces) */
+    traceDir?: string;
+    /** Auth token for remote APIs */
+    authToken?: string;
+}
+interface NormalizedSpan {
+    traceId: string;
+    spanId: string;
+    parentSpanId?: string;
+    name: string;
+    startTimeMs: number;
+    endTimeMs: number;
+    durationMs: number;
+    attributes: Record<string, string | number | boolean>;
+    events: Array<{
+        name: string;
+        attributes: Record<string, string | number | boolean>;
+    }>;
+    status: 'ok' | 'error';
+}
+interface SpanTree {
+    span: NormalizedSpan;
+    children: SpanTree[];
+}
+/**
+ * Fetch trace spans by trace ID. Tries backends in order:
+ * 1. Explicit backend (config.type)
+ * 2. Grafana Tempo (auto-detect)
+ * 3. Jaeger (auto-detect)
+ * 4. Local NDJSON files
+ */
+export declare function fetchTraceSpans(traceId: string, config?: Partial<TraceBackendConfig>): Promise<NormalizedSpan[]>;
+/**
+ * Find the NDJSON trace file matching a given trace ID.
+ * Scans the first line of each `.ndjson` file in the trace directory.
+ */
+export declare function findTraceFile(traceId: string, traceDir?: string): Promise<string | null>;
+export interface RenderTreeOptions {
+    maxDepth?: number;
+    maxChars?: number;
+    /** Fallback intent text for AI spans whose input is too truncated */
+    fallbackIntent?: string;
+    /** When true, don't truncate output/intent previews */
+    fullOutput?: boolean;
+    /** Final task response from the task store (full, not OTEL-truncated) */
+    taskResponse?: string;
+}
+/**
+ * Serialize a trace (by ID) into a compact text tree for LLM prompts.
+ * Auto-detects the backend (Grafana Tempo, Jaeger, local files).
+ */
+export declare function serializeTraceForPrompt(traceIdOrPath: string, maxChars?: number, backendConfig?: Partial<TraceBackendConfig>, 
+/** Final task response from the task store (not truncated by OTEL) */
+taskResponse?: string): Promise<string>;
+/**
+ * Render a span tree as a compact ASCII tree with durations and details.
+ */
+export declare function renderSpanTree(tree: SpanTree, opts?: RenderTreeOptions): string;
+export declare function renderTraceTree(tree: any, // ExecutionNode from trace-reader
+opts?: RenderTreeOptions): string;
+/**
+ * Render a span tree as YAML-like structured output.
+ * More readable than ASCII tree for JSON inputs/outputs.
+ */
+export declare function renderSpanYaml(tree: SpanTree, allSpans: NormalizedSpan[], opts?: RenderTreeOptions): string;
+export {};
+//# sourceMappingURL=trace-serializer.d.ts.map

package/dist/agent-protocol/trace-serializer.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"trace-serializer.d.ts","sourceRoot":"","sources":["file:///home/runner/work/visor/visor/src/agent-protocol/trace-serializer.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;GAcG;AAWH,MAAM,WAAW,kBAAkB;IACjC,IAAI,EAAE,SAAS,GAAG,QAAQ,GAAG,MAAM,GAAG,MAAM,CAAC;IAC7C,mDAAmD;IACnD,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,yDAAyD;IACzD,mBAAmB,CAAC,EAAE,MAAM,GAAG,MAAM,CAAC;IACtC,mDAAmD;IACnD,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,qDAAqD;IACrD,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,iCAAiC;IACjC,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB;AAmBD,UAAU,cAAc;IACtB,OAAO,EAAE,MAAM,CAAC;IAChB,MAAM,EAAE,MAAM,CAAC;IACf,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,IAAI,EAAE,MAAM,CAAC;IACb,WAAW,EAAE,MAAM,CAAC;IACpB,SAAS,EAAE,MAAM,CAAC;IAClB,UAAU,EAAE,MAAM,CAAC;IACnB,UAAU,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,GAAG,MAAM,GAAG,OAAO,CAAC,CAAC;IACtD,MAAM,EAAE,KAAK,CAAC;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,UAAU,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,GAAG,MAAM,GAAG,OAAO,CAAC,CAAA;KAAE,CAAC,CAAC;IACvF,MAAM,EAAE,IAAI,GAAG,OAAO,CAAC;CACxB;AAED,UAAU,QAAQ;IAChB,IAAI,EAAE,cAAc,CAAC;IACrB,QAAQ,EAAE,QAAQ,EAAE,CAAC;CACtB;AAqKD;;;;;;GAMG;AACH,wBAAsB,eAAe,CACnC,OAAO,EAAE,MAAM,EACf,MAAM,CAAC,EAAE,OAAO,CAAC,kBAAkB,CAAC,GACnC,OAAO,CAAC,cAAc,EAAE,CAAC,CA0B3B;AA2ID;;;GAGG;AACH,wBAAsB,aAAa,CAAC,OAAO,EAAE,MAAM,EAAE,QAAQ,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC,MAAM,GAAG,IAAI,CAAC,CAmB9F;AAgJD,MAAM,WAAW,iBAAiB;IAChC,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,qEAAqE;IACrE,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,uDAAuD;IACvD,UAAU,CAAC,EAAE,OAAO,CAAC;IACrB,yEAAyE;IACzE,YAAY,CAAC,EAAE,MAAM,CAAC;CACvB;AA4CD;;;GAGG;AACH,wBAAsB,uBAAuB,CAC3C,aAAa,EAAE,MAAM,EACrB,QAAQ,CAAC,EAAE,MAAM,EACjB,aAAa,CAAC,EAAE,OAAO,CAAC,kBAAkB,CAAC;AAC3C,sEAAsE;AACtE,YAAY,CAAC,EAAE,MAAM,GACpB,OAAO,CAAC,MAAM,CAAC,CA+BjB;AAED;;GAEG;AACH,wBAAgB,cAAc,CAAC,IAAI,EAAE,QAAQ,EAAE,IAAI,CAAC,EAAE,iBAAiB,GAAG,MAAM,CAwB/E;AAGD,wBAAgB,eAAe,CAC7B,IAAI,EAAE,GAAG,EAAE,kCAAkC;AAC7C,IAAI,CAAC,EAAE,iBAAiB,GACvB,MAAM,CAqBR;AAMD;;;GAGG;AACH,wBAAgB,cAAc,CAC5B,IAAI,EAAE,QAAQ,EACd,QAAQ,EAAE,cAAc,EAAE,EAC1B,IAAI,CAAC,EAAE,iBAAiB,GACvB,MAAM,CA+BR"}

package/dist/agent-protocol/track-execution.d.ts

@@ -19,6 +19,8 @@ export interface TrackExecutionOptions {
     metadata?: Record<string, unknown>;
     /** Human-readable description of what's being executed */
     messageText: string;
+    /** Run LLM evaluation after task completion (non-blocking). */
+    autoEvaluate?: boolean;
 }
 /**
  * Wrap an engine execution call with task lifecycle tracking.

package/dist/agent-protocol/track-execution.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"track-execution.d.ts","sourceRoot":"","sources":["file:///home/runner/work/visor/visor/src/agent-protocol/track-execution.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AAOH,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,cAAc,CAAC;AAC9C,OAAO,KAAK,EAAgB,SAAS,EAAE,MAAM,SAAS,CAAC;AAEvD,MAAM,MAAM,UAAU,GAClB,KAAK,GACL,OAAO,GACP,UAAU,GACV,OAAO,GACP,UAAU,GACV,OAAO,GACP,KAAK,GACL,KAAK,GACL,KAAK,GACL,WAAW,GACX,SAAS,CAAC;AAEd,MAAM,WAAW,qBAAqB;IACpC,SAAS,EAAE,SAAS,CAAC;IACrB,MAAM,EAAE,UAAU,CAAC;IACnB,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,6EAA6E;IAC7E,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,QAAQ,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IACnC,0DAA0D;IAC1D,WAAW,EAAE,MAAM,CAAC;CACrB;AAED;;;;;;GAMG;AACH,wBAAsB,cAAc,CAAC,CAAC,EACpC,IAAI,EAAE,qBAAqB,EAC3B,QAAQ,EAAE,MAAM,OAAO,CAAC,CAAC,CAAC,GACzB,OAAO,CAAC;IAAE,IAAI,EAAE,SAAS,CAAC;IAAC,MAAM,EAAE,CAAC,CAAA;CAAE,CAAC,CAwFzC"}
+{"version":3,"file":"track-execution.d.ts","sourceRoot":"","sources":["file:///home/runner/work/visor/visor/src/agent-protocol/track-execution.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AAOH,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,cAAc,CAAC;AAC9C,OAAO,KAAK,EAAgB,SAAS,EAAE,MAAM,SAAS,CAAC;AAUvD,MAAM,MAAM,UAAU,GAClB,KAAK,GACL,OAAO,GACP,UAAU,GACV,OAAO,GACP,UAAU,GACV,OAAO,GACP,KAAK,GACL,KAAK,GACL,KAAK,GACL,WAAW,GACX,SAAS,CAAC;AAEd,MAAM,WAAW,qBAAqB;IACpC,SAAS,EAAE,SAAS,CAAC;IACrB,MAAM,EAAE,UAAU,CAAC;IACnB,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,6EAA6E;IAC7E,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,QAAQ,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IACnC,0DAA0D;IAC1D,WAAW,EAAE,MAAM,CAAC;IACpB,+DAA+D;IAC/D,YAAY,CAAC,EAAE,OAAO,CAAC;CACxB;AAED;;;;;;GAMG;AACH,wBAAsB,cAAc,CAAC,CAAC,EACpC,IAAI,EAAE,qBAAqB,EAC3B,QAAQ,EAAE,MAAM,OAAO,CAAC,CAAC,CAAC,GACzB,OAAO,CAAC;IAAE,IAAI,EAAE,SAAS,CAAC;IAAC,MAAM,EAAE,CAAC,CAAA;CAAE,CAAC,CA4GzC"}

package/dist/cli-main.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"","sourceRoot":"","sources":["file:///home/runner/work/visor/visor/src/cli-main.ts"],"names":[],"mappings":"AA2qCA;;GAEG;AACH,wBAAsB,IAAI,IAAI,OAAO,CAAC,IAAI,CAAC,CAmiD1C"}
+{"version":3,"file":"","sourceRoot":"","sources":["file:///home/runner/work/visor/visor/src/cli-main.ts"],"names":[],"mappings":"AA2qCA;;GAEG;AACH,wBAAsB,IAAI,IAAI,OAAO,CAAC,IAAI,CAAC,CAskD1C"}