@gonzih/cc-agent 0.6.0 → 0.7.1

package/dist/docker.js

@@ -0,0 +1,165 @@
+import { execFile, spawn } from "child_process";
+import { EventEmitter } from "events";
+import { promisify } from "util";
+import { logger } from "./logger.js";
+const execFileAsync = promisify(execFile);
+export async function isDockerAvailable() {
+    try {
+        await execFileAsync("docker", ["info"], { timeout: 5000 });
+        return true;
+    }
+    catch {
+        return false;
+    }
+}
+export async function listCcAgentContainers() {
+    try {
+        const { stdout } = await execFileAsync("docker", [
+            "ps",
+            "--filter", "name=cc-agent-",
+            "--format", "{{.ID}}\t{{.Names}}\t{{.Status}}\t{{.RunningFor}}",
+        ]);
+        return stdout
+            .trim()
+            .split("\n")
+            .filter(Boolean)
+            .map((line) => {
+            const [id, name, status, uptime] = line.split("\t");
+            return { id: id ?? "", name: name ?? "", status: status ?? "", uptime: uptime ?? "" };
+        });
+    }
+    catch {
+        return [];
+    }
+}
+/**
+ * Run a cc-agent job inside a Docker container.
+ *
+ * Emits:
+ *   "text"  (line: string)      — each line of container output
+ *   "exit"  (code: number)      — container exit code
+ *   "error" (err: unknown)      — fatal error before container starts
+ */
+export function runDockerAgent(opts) {
+    const emitter = new EventEmitter();
+    emitter.pid = undefined;
+    emitter.stdin = null;
+    let containerStarted = false;
+    let killed = false;
+    emitter.kill = () => {
+        killed = true;
+        if (containerStarted) {
+            execFile("docker", ["rm", "-f", opts.containerName], () => { });
+        }
+    };
+    void (async () => {
+        try {
+            // Build docker env args
+            const envArgs = [];
+            if (opts.anthropicToken) {
+                envArgs.push("-e", `ANTHROPIC_AUTH_TOKEN=${opts.anthropicToken}`);
+                envArgs.push("-e", `ANTHROPIC_API_KEY=${opts.anthropicToken}`);
+            }
+            if (opts.githubToken) {
+                envArgs.push("-e", `GITHUB_TOKEN=${opts.githubToken}`);
+                envArgs.push("-e", `GH_TOKEN=${opts.githubToken}`);
+            }
+            if (opts.namespace) {
+                envArgs.push("-e", `CC_AGENT_NAMESPACE=${opts.namespace}`);
+            }
+            envArgs.push("-e", "HOME=/root");
+            envArgs.push("-e", "GIT_CONFIG_GLOBAL=/dev/null");
+            // Pass task and repo via env to avoid shell quoting issues
+            envArgs.push("-e", `CC_DOCKER_TASK=${opts.task}`);
+            envArgs.push("-e", `CC_DOCKER_REPO=${opts.repoUrl}`);
+            const containerScript = [
+                "set -e",
+                // Install system deps (node:22 is Debian-based)
+                "apt-get update -qq >/dev/null 2>&1 && apt-get install -y -qq git curl >/dev/null 2>&1",
+                // Install gh CLI via direct binary download (amd64)
+                "GH_VERSION=2.65.0",
+                "ARCH=$(dpkg --print-architecture 2>/dev/null || echo amd64)",
+                "curl -fsSL \"https://github.com/cli/cli/releases/download/v${GH_VERSION}/gh_${GH_VERSION}_linux_${ARCH}.tar.gz\" -o /tmp/gh.tar.gz",
+                "tar -xzf /tmp/gh.tar.gz -C /tmp",
+                "mv /tmp/gh_${GH_VERSION}_linux_${ARCH}/bin/gh /usr/local/bin/",
+                // Install claude-code
+                "npm install -g @anthropic-ai/claude-code >/dev/null 2>&1",
+                // Configure git
+                "git config --global user.email 'cc-agent@localhost'",
+                "git config --global user.name 'cc-agent'",
+                // Configure HTTPS credential helper for GitHub token
+                "git config --global credential.helper '!f() { echo username=x-access-token; echo password=$GITHUB_TOKEN; }; f'",
+                // Clone repo
+                "git clone --depth 1 \"$CC_DOCKER_REPO\" /workspace",
+                "cd /workspace",
+                // Run Claude (dangerously-skip-permissions needed for non-interactive use)
+                "exec claude --dangerously-skip-permissions --print --output-format stream-json -p \"$CC_DOCKER_TASK\"",
+            ].join(" && ");
+            if (killed)
+                return;
+            // Start container in detached mode
+            const { stdout: dockerIdRaw } = await execFileAsync("docker", [
+                "run", "-d",
+                "--name", opts.containerName,
+                ...envArgs,
+                "node:22",
+                "/bin/sh", "-c", containerScript,
+            ]);
+            const dockerId = dockerIdRaw.trim();
+            containerStarted = true;
+            logger.info("docker:container-started", { name: opts.containerName, id: dockerId });
+            if (killed) {
+                execFile("docker", ["rm", "-f", opts.containerName], () => { });
+                emitter.emit("exit", 1);
+                return;
+            }
+            // Stream logs from container
+            const logProc = spawn("docker", ["logs", "-f", opts.containerName], {
+                stdio: ["ignore", "pipe", "pipe"],
+            });
+            let buf = "";
+            const onData = (data) => {
+                buf += data.toString();
+                const lines = buf.split("\n");
+                buf = lines.pop() ?? "";
+                for (const line of lines) {
+                    emitter.emit("text", line);
+                }
+            };
+            logProc.stdout?.on("data", onData);
+            logProc.stderr?.on("data", onData);
+            // Wait for container to finish
+            let exitCode = 0;
+            try {
+                const { stdout: waitOut } = await execFileAsync("docker", ["wait", opts.containerName]);
+                exitCode = parseInt(waitOut.trim(), 10);
+                if (isNaN(exitCode))
+                    exitCode = 0;
+            }
+            catch {
+                exitCode = 1;
+            }
+            // Drain remaining buffered output
+            if (buf.trim())
+                emitter.emit("text", buf);
+            logProc.kill();
+            // Cleanup container
+            containerStarted = false;
+            try {
+                await execFileAsync("docker", ["rm", "-f", opts.containerName]);
+            }
+            catch {
+                // Best-effort cleanup
+            }
+            logger.info("docker:container-done", { name: opts.containerName, exitCode });
+            emitter.emit("exit", exitCode);
+        }
+        catch (err) {
+            logger.error("docker:error", { name: opts.containerName, error: String(err) });
+            emitter.emit("error", err);
+            emitter.emit("exit", 1);
+        }
+    })();
+    return emitter;
+}
+//# sourceMappingURL=docker.js.map

package/dist/docker.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"docker.js","sourceRoot":"","sources":["../src/docker.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,QAAQ,EAAE,KAAK,EAAE,MAAM,eAAe,CAAC;AAChD,OAAO,EAAE,YAAY,EAAE,MAAM,QAAQ,CAAC;AACtC,OAAO,EAAE,SAAS,EAAE,MAAM,MAAM,CAAC;AACjC,OAAO,EAAE,MAAM,EAAE,MAAM,aAAa,CAAC;AAErC,MAAM,aAAa,GAAG,SAAS,CAAC,QAAQ,CAAC,CAAC;AAE1C,MAAM,CAAC,KAAK,UAAU,iBAAiB;IACrC,IAAI,CAAC;QACH,MAAM,aAAa,CAAC,QAAQ,EAAE,CAAC,MAAM,CAAC,EAAE,EAAE,OAAO,EAAE,IAAI,EAAyC,CAAC,CAAC;QAClG,OAAO,IAAI,CAAC;IACd,CAAC;IAAC,MAAM,CAAC;QACP,OAAO,KAAK,CAAC;IACf,CAAC;AACH,CAAC;AASD,MAAM,CAAC,KAAK,UAAU,qBAAqB;IACzC,IAAI,CAAC;QACH,MAAM,EAAE,MAAM,EAAE,GAAG,MAAM,aAAa,CAAC,QAAQ,EAAE;YAC/C,IAAI;YACJ,UAAU,EAAE,gBAAgB;YAC5B,UAAU,EAAE,mDAAmD;SAChE,CAAC,CAAC;QACH,OAAO,MAAM;aACV,IAAI,EAAE;aACN,KAAK,CAAC,IAAI,CAAC;aACX,MAAM,CAAC,OAAO,CAAC;aACf,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE;YACZ,MAAM,CAAC,EAAE,EAAE,IAAI,EAAE,MAAM,EAAE,MAAM,CAAC,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;YACpD,OAAO,EAAE,EAAE,EAAE,EAAE,IAAI,EAAE,EAAE,IAAI,EAAE,IAAI,IAAI,EAAE,EAAE,MAAM,EAAE,MAAM,IAAI,EAAE,EAAE,MAAM,EAAE,MAAM,IAAI,EAAE,EAAE,CAAC;QACxF,CAAC,CAAC,CAAC;IACP,CAAC;IAAC,MAAM,CAAC;QACP,OAAO,EAAE,CAAC;IACZ,CAAC;AACH,CAAC;AAQD;;;;;;;GAOG;AACH,MAAM,UAAU,cAAc,CAAC,IAO9B;IACC,MAAM,OAAO,GAAG,IAAI,YAAY,EAAwB,CAAC;IACzD,OAAO,CAAC,GAAG,GAAG,SAAS,CAAC;IACxB,OAAO,CAAC,KAAK,GAAG,IAAI,CAAC;IAErB,IAAI,gBAAgB,GAAG,KAAK,CAAC;IAC7B,IAAI,MAAM,GAAG,KAAK,CAAC;IAEnB,OAAO,CAAC,IAAI,GAAG,GAAG,EAAE;QAClB,MAAM,GAAG,IAAI,CAAC;QACd,IAAI,gBAAgB,EAAE,CAAC;YACrB,QAAQ,CAAC,QAAQ,EAAE,CAAC,IAAI,EAAE,IAAI,EAAE,IAAI,CAAC,aAAa,CAAC,EAAE,GAAG,EAAE,GAAE,CAAC,CAAC,CAAC;QACjE,CAAC;IACH,CAAC,CAAC;IAEF,KAAK,CAAC,KAAK,IAAI,EAAE;QACf,IAAI,CAAC;YACH,wBAAwB;YACxB,MAAM,OAAO,GAAa,EAAE,CAAC;YAC7B,IAAI,IAAI,CAAC,cAAc,EAAE,CAAC;gBACxB,OAAO,CAAC,IAAI,CAAC,IAAI,EAAE,wBAAwB,IAAI,CAAC,cAAc,EAAE,CAAC,CAAC;gBAClE,OAAO,CAAC,IAAI,CAAC,IAAI,EAAE,qBAAqB,IAAI,CAAC,cAAc,EAAE,CAAC,CAAC;YACjE,CAAC;YACD,IAAI,IAAI,CAAC,WAAW,EAAE,CAAC;gBACrB,OAAO,CAAC,IAAI,CAAC,IAAI,EAAE,gBAAgB,IAAI,CAAC,WAAW,EAAE,CAAC,CAAC;gBACvD,OAAO,CAAC,IAAI,CAAC,IAAI,EAAE,YAAY,IAAI,CAAC,WAAW,EAAE,CAAC,CAAC;YACrD,CAAC;YACD,IAAI,IAAI,CAAC,SAAS,EAAE,CAAC;gBACnB,OAAO,CAAC,IAAI,CAAC,IAAI,EAAE,sBAAsB,IAAI,CAAC,SAAS,EAAE,CAAC,CAAC;YAC7D,CAAC;YACD,OAAO,CAAC,IAAI,CAAC,IAAI,EAAE,YAAY,CAAC,CAAC;YACjC,OAAO,CAAC,IAAI,CAAC,IAAI,EAAE,6BAA6B,CAAC,CAAC;YAClD,2DAA2D;YAC3D,OAAO,CAAC,IAAI,CAAC,IAAI,EAAE,kBAAkB,IAAI,CAAC,IAAI,EAAE,CAAC,CAAC;YAClD,OAAO,CAAC,IAAI,CAAC,IAAI,EAAE,kBAAkB,IAAI,CAAC,OAAO,EAAE,CAAC,CAAC;YAErD,MAAM,eAAe,GAAG;gBACtB,QAAQ;gBACR,gDAAgD;gBAChD,uFAAuF;gBACvF,oDAAoD;gBACpD,mBAAmB;gBACnB,6DAA6D;gBAC7D,oIAAoI;gBACpI,iCAAiC;gBACjC,+DAA+D;gBAC/D,sBAAsB;gBACtB,0DAA0D;gBAC1D,gBAAgB;gBAChB,qDAAqD;gBACrD,0CAA0C;gBAC1C,qDAAqD;gBACrD,gHAAgH;gBAChH,aAAa;gBACb,oDAAoD;gBACpD,eAAe;gBACf,2EAA2E;gBAC3E,uGAAuG;aACxG,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;YAEf,IAAI,MAAM;gBAAE,OAAO;YAEnB,mCAAmC;YACnC,MAAM,EAAE,MAAM,EAAE,WAAW,EAAE,GAAG,MAAM,aAAa,CAAC,QAAQ,EAAE;gBAC5D,KAAK,EAAE,IAAI;gBACX,QAAQ,EAAE,IAAI,CAAC,aAAa;gBAC5B,GAAG,OAAO;gBACV,SAAS;gBACT,SAAS,EAAE,IAAI,EAAE,eAAe;aACjC,CAAC,CAAC;YACH,MAAM,QAAQ,GAAG,WAAW,CAAC,IAAI,EAAE,CAAC;YACpC,gBAAgB,GAAG,IAAI,CAAC;YACxB,MAAM,CAAC,IAAI,CAAC,0BAA0B,EAAE,EAAE,IAAI,EAAE,IAAI,CAAC,aAAa,EAAE,EAAE,EAAE,QAAQ,EAAE,CAAC,CAAC;YAEpF,IAAI,MAAM,EAAE,CAAC;gBACX,QAAQ,CAAC,QAAQ,EAAE,CAAC,IAAI,EAAE,IAAI,EAAE,IAAI,CAAC,aAAa,CAAC,EAAE,GAAG,EAAE,GAAE,CAAC,CAAC,CAAC;gBAC/D,OAAO,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;gBACxB,OAAO;YACT,CAAC;YAED,6BAA6B;YAC7B,MAAM,OAAO,GAAG,KAAK,CAAC,QAAQ,EAAE,CAAC,MAAM,EAAE,IAAI,EAAE,IAAI,CAAC,aAAa,CAAC,EAAE;gBAClE,KAAK,EAAE,CAAC,QAAQ,EAAE,MAAM,EAAE,MAAM,CAAC;aAClC,CAAC,CAAC;YAEH,IAAI,GAAG,GAAG,EAAE,CAAC;YACb,MAAM,MAAM,GAAG,CAAC,IAAY,EAAQ,EAAE;gBACpC,GAAG,IAAI,IAAI,CAAC,QAAQ,EAAE,CAAC;gBACvB,MAAM,KAAK,GAAG,GAAG,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;gBAC9B,GAAG,GAAG,KAAK,CAAC,GAAG,EAAE,IAAI,EAAE,CAAC;gBACxB,KAAK,MAAM,IAAI,IAAI,KAAK,EAAE,CAAC;oBACzB,OAAO,CAAC,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,CAAC;gBAC7B,CAAC;YACH,CAAC,CAAC;YACF,OAAO,CAAC,MAAM,EAAE,EAAE,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;YACnC,OAAO,CAAC,MAAM,EAAE,EAAE,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;YAEnC,+BAA+B;YAC/B,IAAI,QAAQ,GAAG,CAAC,CAAC;YACjB,IAAI,CAAC;gBACH,MAAM,EAAE,MAAM,EAAE,OAAO,EAAE,GAAG,MAAM,aAAa,CAAC,QAAQ,EAAE,CAAC,MAAM,EAAE,IAAI,CAAC,aAAa,CAAC,CAAC,CAAC;gBACxF,QAAQ,GAAG,QAAQ,CAAC,OAAO,CAAC,IAAI,EAAE,EAAE,EAAE,CAAC,CAAC;gBACxC,IAAI,KAAK,CAAC,QAAQ,CAAC;oBAAE,QAAQ,GAAG,CAAC,CAAC;YACpC,CAAC;YAAC,MAAM,CAAC;gBACP,QAAQ,GAAG,CAAC,CAAC;YACf,CAAC;YAED,kCAAkC;YAClC,IAAI,GAAG,CAAC,IAAI,EAAE;gBAAE,OAAO,CAAC,IAAI,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC;YAC1C,OAAO,CAAC,IAAI,EAAE,CAAC;YAEf,oBAAoB;YACpB,gBAAgB,GAAG,KAAK,CAAC;YACzB,IAAI,CAAC;gBACH,MAAM,aAAa,CAAC,QAAQ,EAAE,CAAC,IAAI,EAAE,IAAI,EAAE,IAAI,CAAC,aAAa,CAAC,CAAC,CAAC;YAClE,CAAC;YAAC,MAAM,CAAC;gBACP,sBAAsB;YACxB,CAAC;YAED,MAAM,CAAC,IAAI,CAAC,uBAAuB,EAAE,EAAE,IAAI,EAAE,IAAI,CAAC,aAAa,EAAE,QAAQ,EAAE,CAAC,CAAC;YAC7E,OAAO,CAAC,IAAI,CAAC,MAAM,EAAE,QAAQ,CAAC,CAAC;QACjC,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,MAAM,CAAC,KAAK,CAAC,cAAc,EAAE,EAAE,IAAI,EAAE,IAAI,CAAC,aAAa,EAAE,KAAK,EAAE,MAAM,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC;YAC/E,OAAO,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,CAAC,CAAC;YAC3B,OAAO,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;QAC1B,CAAC;IACH,CAAC,CAAC,EAAE,CAAC;IAEL,OAAO,OAAO,CAAC;AACjB,CAAC"}

package/dist/evaluator.d.ts

@@ -0,0 +1,14 @@
+/**
+ * Evaluator job template — generates task text for evaluator jobs in evolutionary plans.
+ */
+export type BranchEval = "test_pass_rate" | "pr_merged" | "manual";
+export type BranchSelect = "best_score" | "score_prop" | "latest";
+export interface EvaluatorOptions {
+    variantJobIds: string[];
+    variantBranches: (string | undefined)[];
+    branchEval: BranchEval;
+    branchSelect: BranchSelect;
+    stepId: string;
+}
+export declare function buildEvaluatorTask(opts: EvaluatorOptions): string;
+//# sourceMappingURL=evaluator.d.ts.map

package/dist/evaluator.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"evaluator.d.ts","sourceRoot":"","sources":["../src/evaluator.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,MAAM,MAAM,UAAU,GAAG,gBAAgB,GAAG,WAAW,GAAG,QAAQ,CAAC;AACnE,MAAM,MAAM,YAAY,GAAG,YAAY,GAAG,YAAY,GAAG,QAAQ,CAAC;AAElE,MAAM,WAAW,gBAAgB;IAC/B,aAAa,EAAE,MAAM,EAAE,CAAC;IACxB,eAAe,EAAE,CAAC,MAAM,GAAG,SAAS,CAAC,EAAE,CAAC;IACxC,UAAU,EAAE,UAAU,CAAC;IACvB,YAAY,EAAE,YAAY,CAAC;IAC3B,MAAM,EAAE,MAAM,CAAC;CAChB;AAED,wBAAgB,kBAAkB,CAAC,IAAI,EAAE,gBAAgB,GAAG,MAAM,CAmDjE"}

package/dist/evaluator.js

@@ -0,0 +1,88 @@
+/**
+ * Evaluator job template — generates task text for evaluator jobs in evolutionary plans.
+ */
+export function buildEvaluatorTask(opts) {
+    const { variantJobIds, variantBranches, branchEval, branchSelect, stepId } = opts;
+    const variantList = variantJobIds
+        .map((id, i) => `  - Variant ${i + 1}: job_id=${id}${variantBranches[i] ? `, branch=${variantBranches[i]}` : ""}`)
+        .join("\n");
+    const evalInstructions = buildEvalInstructions(branchEval);
+    const selectInstructions = buildSelectInstructions(branchSelect, variantJobIds.length);
+    return `You are an evaluator agent for an evolutionary branching plan (step: ${stepId}).
+
+Your job is to evaluate ${variantJobIds.length} variant solutions, score them, and select the best one.
+
+## Variants to Evaluate
+
+${variantList}
+
+## Evaluation Instructions
+
+${evalInstructions}
+
+## Scoring
+
+For each variant, compute a score from 0.0 to 1.0:
+- Check the job output using get_job_output for each variant job_id
+- ${branchEval === "test_pass_rate" ? "Parse test results: look for patterns like 'X passing', 'X tests passed', 'X failed'. Score = (passing / (passing + failing)) * 0.7 + (exitCode === 0 ? 0.3 : 0)" : ""}
+- ${branchEval === "pr_merged" ? "Check if a PR was merged: score = pr_merged ? 1.0 : (pr_exists ? 0.5 : 0.0)" : ""}
+- ${branchEval === "manual" ? "Review the output quality manually and assign a score from 0.0 to 1.0 based on completeness and correctness" : ""}
+- If a variant failed (status=failed or non-zero exit), score it 0.0
+
+After computing each score, call set_job_score with the job_id and computed score.
+
+## Winner Selection
+
+${selectInstructions}
+
+## Output
+
+After evaluating all variants and calling set_job_score for each, output a JSON block exactly like this (on its own line):
+
+WINNER: {"job_id": "<winning_job_id>", "variant_index": <N>, "branch": "<branch_or_null>", "score": <score>, "reason": "<brief reason>"}
+
+This line will be parsed by downstream jobs to know which variant won.
+
+## Important Notes
+
+- Always call set_job_score for ALL variants, even if they scored 0.0
+- Be objective in your evaluation
+- If all variants scored 0.0, pick the one with the least errors or pick variant 1
+`;
+}
+function buildEvalInstructions(branchEval) {
+    switch (branchEval) {
+        case "test_pass_rate":
+            return `For each variant job:
+1. Call get_job_output with the variant's job_id
+2. Search output for test result patterns: "X passing", "X tests passed", "X failed", "X failures"
+3. Calculate pass rate = passing_tests / (passing_tests + failing_tests)
+4. Check job exit code (exitCode=0 means success, non-zero means failure)
+5. Score = pass_rate * 0.7 + (exitCode === 0 ? 0.3 : 0)`;
+        case "pr_merged":
+            return `For each variant job:
+1. Call get_job_output with the variant's job_id to find the PR URL
+2. Check if the PR was merged by looking for "merged" status in the output
+3. Score = pr_merged ? 1.0 : (pr_created ? 0.5 : 0.0)`;
+        case "manual":
+            return `For each variant job:
+1. Call get_job_output with the variant's job_id to review the full output
+2. Assess the quality, completeness, and correctness of the work
+3. Assign a score from 0.0 to 1.0 based on your assessment`;
+    }
+}
+function buildSelectInstructions(branchSelect, variantCount) {
+    switch (branchSelect) {
+        case "best_score":
+            return `Select the variant with the highest score. If there is a tie, pick the lowest variant index.`;
+        case "score_prop":
+            return `Select a winner using score-proportional (roulette wheel) selection:
+1. Compute selection probability for each variant: p_i = score_i / sum(all_scores)
+2. If all scores are 0, use uniform probability (1/${variantCount} each)
+3. Generate a random number between 0 and 1, then pick the variant whose cumulative probability bracket contains that number
+4. Higher score = more likely to be selected, but lower scorers can still win (prevents premature convergence)`;
+        case "latest":
+            return `Select the variant with the most recent completion time (last to finish). If unsure, pick the highest variant index.`;
+    }
+}
+//# sourceMappingURL=evaluator.js.map

package/dist/evaluator.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"evaluator.js","sourceRoot":"","sources":["../src/evaluator.ts"],"names":[],"mappings":"AAAA;;GAEG;AAaH,MAAM,UAAU,kBAAkB,CAAC,IAAsB;IACvD,MAAM,EAAE,aAAa,EAAE,eAAe,EAAE,UAAU,EAAE,YAAY,EAAE,MAAM,EAAE,GAAG,IAAI,CAAC;IAElF,MAAM,WAAW,GAAG,aAAa;SAC9B,GAAG,CAAC,CAAC,EAAE,EAAE,CAAC,EAAE,EAAE,CAAC,eAAe,CAAC,GAAG,CAAC,YAAY,EAAE,GAAG,eAAe,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,YAAY,eAAe,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC;SACjH,IAAI,CAAC,IAAI,CAAC,CAAC;IAEd,MAAM,gBAAgB,GAAG,qBAAqB,CAAC,UAAU,CAAC,CAAC;IAC3D,MAAM,kBAAkB,GAAG,uBAAuB,CAAC,YAAY,EAAE,aAAa,CAAC,MAAM,CAAC,CAAC;IAEvF,OAAO,wEAAwE,MAAM;;0BAE7D,aAAa,CAAC,MAAM;;;;EAI5C,WAAW;;;;EAIX,gBAAgB;;;;;;IAMd,UAAU,KAAK,gBAAgB,CAAC,CAAC,CAAC,kKAAkK,CAAC,CAAC,CAAC,EAAE;IACzM,UAAU,KAAK,WAAW,CAAC,CAAC,CAAC,6EAA6E,CAAC,CAAC,CAAC,EAAE;IAC/G,UAAU,KAAK,QAAQ,CAAC,CAAC,CAAC,6GAA6G,CAAC,CAAC,CAAC,EAAE;;;;;;;EAO9I,kBAAkB;;;;;;;;;;;;;;;CAenB,CAAC;AACF,CAAC;AAED,SAAS,qBAAqB,CAAC,UAAsB;IACnD,QAAQ,UAAU,EAAE,CAAC;QACnB,KAAK,gBAAgB;YACnB,OAAO;;;;;wDAK2C,CAAC;QAErD,KAAK,WAAW;YACd,OAAO;;;sDAGyC,CAAC;QAEnD,KAAK,QAAQ;YACX,OAAO;;;2DAG8C,CAAC;IAC1D,CAAC;AACH,CAAC;AAED,SAAS,uBAAuB,CAAC,YAA0B,EAAE,YAAoB;IAC/E,QAAQ,YAAY,EAAE,CAAC;QACrB,KAAK,YAAY;YACf,OAAO,8FAA8F,CAAC;QAExG,KAAK,YAAY;YACf,OAAO;;qDAEwC,YAAY;;+GAE8C,CAAC;QAE5G,KAAK,QAAQ;YACX,OAAO,sHAAsH,CAAC;IAClI,CAAC;AACH,CAAC"}

package/dist/index.js

@@ -20,11 +20,13 @@ import { Server } from "@modelcontextprotocol/sdk/server/index.js";
 import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
 import { CallToolRequestSchema, ListToolsRequestSchema, } from "@modelcontextprotocol/sdk/types.js";
 import { JobManager } from "./agent.js";
+import { buildEvaluatorTask } from "./evaluator.js";
 import { loadProfiles, upsertProfile, deleteProfile, getProfile, interpolate } from "./profiles.js";
 import { planStore, jobStore, learningsStore } from "./store.js";
 import { getNamespace } from "./namespace.js";
 import { initRedis } from "./redis.js";
 import { logger } from "./logger.js";
+import { listCcAgentContainers } from "./docker.js";
 import { v4 as uuidv4 } from "uuid";
 import { execFile } from "child_process";
 import { promisify } from "util";
@@ -115,6 +117,18 @@ server.setRequestHandler(ListToolsRequestSchema, async () => ({
                         type: "string",
                         description: "Ollama host URL (default: 'http://localhost:11434'). Only used when ollama_model is set.",
                     },
+                    docker_isolation: {
+                        type: "boolean",
+                        description: "Run the agent in a fresh Docker container for full filesystem and process isolation. Requires Docker (colima or Docker Desktop) to be running. Falls back to host mode if Docker is unavailable. Default: false.",
+                    },
+                    smoke_test: {
+                        type: "string",
+                        description: "Shell command to run as a cheap pre-check before the full task. If it exits non-zero or times out, the job fails immediately. Example: 'npm test -- --testPathPattern=smoke 2>&1 | tail -5'",
+                    },
+                    smoke_test_timeout: {
+                        type: "number",
+                        description: "Timeout for the smoke test in seconds (default 60). Only used when smoke_test is set.",
+                    },
                 },
                 required: ["repo_url", "task"],
             },
@@ -148,7 +162,15 @@ server.setRequestHandler(ListToolsRequestSchema, async () => ({
         {
             name: "list_jobs",
             description: "List all agent jobs (running, done, failed, cancelled).",
-            inputSchema: { type: "object", properties: {} },
+            inputSchema: {
+                type: "object",
+                properties: {
+                    min_score: {
+                        type: "number",
+                        description: "Only return jobs with score >= this value (0.0–1.0). Unscored jobs are excluded when this filter is set.",
+                    },
+                },
+            },
         },
         {
             name: "cancel_job",
@@ -280,6 +302,20 @@ server.setRequestHandler(ListToolsRequestSchema, async () => ({
                                     items: { type: "string" },
                                     description: "Step IDs (from this plan) that must complete before this step starts",
                                 },
+                                branches: {
+                                    type: "number",
+                                    description: "If set, spawn this many parallel variant jobs for this step instead of 1. An evaluator job is automatically added to score and select the best variant.",
+                                },
+                                branch_eval: {
+                                    type: "string",
+                                    enum: ["test_pass_rate", "pr_merged", "manual"],
+                                    description: "How to score variants: test_pass_rate (parse test output), pr_merged (check PR status), manual (evaluator uses judgment). Default: test_pass_rate",
+                                },
+                                branch_select: {
+                                    type: "string",
+                                    enum: ["best_score", "score_prop", "latest"],
+                                    description: "How to pick the winner: best_score (highest score wins), score_prop (score-proportional random selection), latest (most recently completed). Default: best_score",
+                                },
                             },
                             required: ["id", "repo_url", "task"],
                         },
@@ -381,6 +417,19 @@ server.setRequestHandler(ListToolsRequestSchema, async () => ({
                 required: ["job_id"],
             },
         },
+        {
+            name: "set_job_score",
+            description: "Set a quality score (0.0–1.0) on a completed job. Used by evaluator agents in evolutionary branching plans to record how well each variant performed.",
+            inputSchema: {
+                type: "object",
+                properties: {
+                    job_id: { type: "string", description: "Job ID to score" },
+                    score: { type: "number", description: "Score from 0.0 to 1.0" },
+                    reason: { type: "string", description: "Optional reason or explanation for the score" },
+                },
+                required: ["job_id", "score"],
+            },
+        },
         {
             name: "get_learnings",
             description: "Return accumulated learnings for a namespace. Learnings are written by agents at the end of each job and stored per-namespace. Use this to understand what prior agents have discovered.",
@@ -411,6 +460,11 @@ server.setRequestHandler(ListToolsRequestSchema, async () => ({
                 },
             },
         },
+        {
+            name: "docker_ps",
+            description: "List currently running cc-agent Docker containers. Shows container name, status, and uptime.",
+            inputSchema: { type: "object", properties: {} },
+        },
         {
             name: "spawn_from_profile",
             description: "Spawn an agent job from a saved profile. Supports variable interpolation and per-call overrides.",
@@ -466,6 +520,9 @@ server.setRequestHandler(CallToolRequestSchema, async (req) => {
                 model: a.model,
                 ollamaModel: a.ollama_model,
                 ollamaHost: a.ollama_host,
+                dockerIsolation: a.docker_isolation,
+                smokeTest: a.smoke_test,
+                smokeTestTimeout: a.smoke_test_timeout,
                 requiresApproval: !isTrusted,
             });
             if (!isTrusted && owner) {
@@ -545,6 +602,8 @@ server.setRequestHandler(CallToolRequestSchema, async (req) => {
                             cost_usd: job.costUsd,
                             usage: job.usage,
                             approval_issue_url: job.approvalIssueUrl,
+                            score: job.score ?? null,
+                            score_source: job.scoreSource ?? null,
                         }),
                     },
                 ],
@@ -572,7 +631,11 @@ server.setRequestHandler(CallToolRequestSchema, async (req) => {
         }
         case "list_jobs": {
             logger.info("tool:list_jobs");
-            const jobs = (await jobStore.listJobs()) ?? [];
+            const minScore = typeof a.min_score === "number" ? a.min_score : undefined;
+            let jobs = (await jobStore.listJobs()) ?? [];
+            if (minScore !== undefined) {
+                jobs = jobs.filter((j) => j.score != null && j.score >= minScore);
+            }
             const namespace = getNamespace();
             const learnings_count = await learningsStore.getLearningsCount(namespace);
             return {
@@ -737,16 +800,72 @@ server.setRequestHandler(CallToolRequestSchema, async (req) => {
                         throw new Error(`Step '${step.id}' depends_on unknown step '${sid}'`);
                     return jobId;
                 });
-                const jobId = await manager.spawn({
-                    repoUrl: step.repo_url,
-                    task: step.task,
-                    createBranch: step.create_branch,
-                    dependsOn: resolvedDeps,
-                });
-                stepIdToJobId.set(step.id, jobId);
-                results.push({ stepId: step.id, jobId, status: resolvedDeps?.length ? "pending" : "cloning" });
+                if (step.branches && step.branches > 1) {
+                    // Evolutionary mode: spawn N variant jobs in parallel
+                    const branchEval = step.branch_eval ?? "test_pass_rate";
+                    const branchSelect = step.branch_select ?? "best_score";
+                    const variantJobIds = [];
+                    const variantBranches = [];
+                    for (let i = 1; i <= step.branches; i++) {
+                        const branchName = step.create_branch ? `${step.create_branch}-v${i}` : undefined;
+                        variantBranches.push(branchName);
+                        const jobId = await manager.spawn({
+                            repoUrl: step.repo_url,
+                            task: step.task,
+                            createBranch: branchName,
+                            dependsOn: resolvedDeps,
+                            variantIndex: i,
+                        });
+                        variantJobIds.push(jobId);
+                    }
+                    // Update siblings on all variant jobs
+                    for (const jobId of variantJobIds) {
+                        manager.setJobSiblings(jobId, variantJobIds.filter((id) => id !== jobId));
+                    }
+                    // Build evaluator task and spawn evaluator job
+                    const evalTask = buildEvaluatorTask({
+                        variantJobIds,
+                        variantBranches,
+                        branchEval,
+                        branchSelect,
+                        stepId: step.id,
+                    });
+                    const evalJobId = await manager.spawn({
+                        repoUrl: step.repo_url,
+                        task: evalTask,
+                        dependsOn: variantJobIds,
+                    });
+                    // The logical step ID maps to the evaluator job (so subsequent steps depend on it)
+                    stepIdToJobId.set(step.id, evalJobId);
+                    // Track variant jobs
+                    for (let i = 0; i < variantJobIds.length; i++) {
+                        results.push({
+                            stepId: `${step.id}-v${i + 1}`,
+                            jobId: variantJobIds[i],
+                            status: resolvedDeps?.length ? "pending" : "cloning",
+                            role: "variant",
+                        });
+                    }
+                    // Track evaluator job
+                    results.push({
+                        stepId: step.id,
+                        jobId: evalJobId,
+                        status: "pending",
+                        role: "evaluator",
+                    });
+                }
+                else {
+                    // Standard single job
+                    const jobId = await manager.spawn({
+                        repoUrl: step.repo_url,
+                        task: step.task,
+                        createBranch: step.create_branch,
+                        dependsOn: resolvedDeps,
+                    });
+                    stepIdToJobId.set(step.id, jobId);
+                    results.push({ stepId: step.id, jobId, status: resolvedDeps?.length ? "pending" : "cloning" });
+                }
             }
-            // Persist the plan record
             const planId = uuidv4();
             planStore.savePlan({ id: planId, goal, steps: results, createdAt: new Date().toISOString() }).catch(() => { });
             return {
@@ -903,6 +1022,13 @@ server.setRequestHandler(CallToolRequestSchema, async (req) => {
                 content: [{ type: "text", text: JSON.stringify(result) }],
             };
         }
+        case "set_job_score": {
+            logger.info("tool:set_job_score", { job_id: a.job_id, score: a.score });
+            const result = manager.setJobScore(a.job_id, a.score, a.reason);
+            return {
+                content: [{ type: "text", text: JSON.stringify(result) }],
+            };
+        }
         case "get_learnings": {
             const ns = a.namespace ?? getNamespace();
             const limit = typeof a.limit === "number" ? a.limit : 10;
@@ -926,6 +1052,16 @@ server.setRequestHandler(CallToolRequestSchema, async (req) => {
                     }],
             };
         }
+        case "docker_ps": {
+            logger.info("tool:docker_ps");
+            const containers = await listCcAgentContainers();
+            return {
+                content: [{
+                        type: "text",
+                        text: JSON.stringify({ containers, total: containers.length }),
+                    }],
+            };
+        }
         default:
             throw new Error(`Unknown tool: ${name}`);
     }