Improve autofix: simplify config, add loop prevention, strengthen Fix mode (#181)

* Improve autofix * UI * remove unused TriggerField props, improve bot commit detection - Remove `alternateEnabledValue` and `enabledContent` props from TriggerField (dead code, not used by any caller) - Move `isBotCommit` to module scope and check both `author.name` and `committer.name` for [bot] suffix * fix: truncate workflow_runs before schema change existing records don't have repoId, causing NOT NULL constraint failure * truncate workflow_runs before adding NOT NULL repoId existing rows don't have repoId values and can't be migrated --------- Co-authored-by: pullfrog[bot] <226033991+pullfrog[bot]@users.noreply.github.com>
2026-01-27 03:54:46 +00:00
parent d545a84027
commit 2514bb1cf7
4 changed files with 557 additions and 169 deletions
@@ -7,7 +7,7 @@ this directory contains the mcp (model context protocol) server tools for intera
 ### check suite tools

 #### `get_check_suite_logs`
-get workflow run logs for a failed check suite.
+get workflow run logs for a failed check suite with intelligent log analysis.

 **parameters:**
 - `check_suite_id` (number): the id from check_suite.id in the webhook payload
@@ -15,16 +15,41 @@ get workflow run logs for a failed check suite.
 **replaces:** `gh run list` and `gh run view --log`

 **returns:**
-all logs from all failed workflow runs in the check suite, including:
- workflow run details (id, name, html_url, conclusion)
- job details for each workflow run (id, name, status, conclusion, logs)
+structured failure information for each failed job:
+- `_instructions`: explains how to use each field
+- `failed_jobs[]`: array of failed job results, each containing:
+  - `job_id`, `job_name`, `job_url`: job identification
+  - `failed_steps`: which CI steps failed (e.g., "Step 6: Run tests")
+  - `log_index`: array of interesting lines (errors, warnings, failures) with line numbers
+  - `excerpt`: ~80 line curated window around the last error
+  - `full_log_path`: path to complete log file for deeper investigation
+
+**log_index types:**
+- `error`: lines matching `##[error]`, `Error:`, `ERR_`, `exit code N`
+- `warning`: lines matching `##[warning]`, `WARN`
+- `failure`: lines matching `N failed`, `FAIL`, `✕`
+- `trace`: stack trace lines (deduplicated)
+
+**workflow for using results:**
+1. scan `log_index` to see where errors/warnings/failures are located in the log
+2. read `excerpt` for immediate context around the main error
+3. if excerpt doesn't show what you need, read specific line ranges from `full_log_path`
+4. check `failed_steps` and read the workflow yml to understand what command failed

 **example:**
 ```typescript
 // when handling a check_suite_completed webhook
-await mcp.call("gh_pullfrog/get_check_suite_logs", {
+const result = await mcp.call("gh_pullfrog/get_check_suite_logs", {
  check_suite_id: check_suite.id
 });
+
+// result.failed_jobs[0].log_index shows:
+// [
+//   { line: 181, content: "WARN  Failed to create bin...", type: "warning" },
+//   { line: 1079, content: "Error: expect(received).toBe(expected)", type: "error" },
+//   ...
+// ]
+// use these line numbers to read specific sections from full_log_path
 ```

 ### review tools
@@ -1,4 +1,7 @@
+import { mkdirSync, writeFileSync } from "node:fs";
+import { join } from "node:path";
 import { type } from "arktype";
+import { log } from "../utils/log.ts";
 import type { ToolContext } from "./server.ts";
 import { execute, tool } from "./shared.ts";

@@ -6,13 +9,127 @@ export const GetCheckSuiteLogs = type({
  check_suite_id: type.number.describe("the id from check_suite.id"),
 });

+type LogLine = {
+  line: number;
+  content: string;
+  type: "error" | "warning" | "failure" | "trace";
+};
+
+type LogAnalysis = {
+  totalLines: number;
+  index: LogLine[];
+  excerpt: {
+    content: string;
+    startLine: number;
+    endLine: number;
+  };
+};
+
+function analyzeLog(logs: string, excerptLines = 80): LogAnalysis {
+  // biome-ignore lint/suspicious/noControlCharactersInRegex: ANSI escape codes use control chars
+  const clean = logs.replace(/\x1b\[[0-9;]*m/g, "");
+  const lines = clean.split("\n");
+  const totalLines = lines.length;
+
+  const index: LogLine[] = [];
+
+  const patterns: Array<{ type: LogLine["type"]; pattern: RegExp; skip?: RegExp }> = [
+    { type: "error", pattern: /##\[error\]/i },
+    { type: "error", pattern: /\bError:/i },
+    { type: "error", pattern: /\bERR_/i },
+    { type: "error", pattern: /exit code [1-9]/i },
+    { type: "warning", pattern: /##\[warning\]/i },
+    { type: "warning", pattern: /\bWARN\b/i, skip: /apt|dpkg|Reading package/i },
+    { type: "failure", pattern: /\d+ failed/i },
+    { type: "failure", pattern: /FAIL\b/i },
+    { type: "failure", pattern: /✕|✗|×/ },
+    { type: "trace", pattern: /^\s+at\s+/i },
+  ];
+
+  for (let i = 0; i < lines.length; i++) {
+    const line = lines[i];
+
+    for (const p of patterns) {
+      if (p.pattern.test(line)) {
+        if (p.skip?.test(line)) continue;
+
+        // dedupe consecutive traces
+        if (p.type === "trace" && index.length > 0 && index[index.length - 1].type === "trace") {
+          continue;
+        }
+
+        // truncate long lines
+        const truncated = line.length > 120 ? line.slice(0, 117) + "..." : line;
+
+        index.push({
+          line: i + 1,
+          content: truncated.trim(),
+          type: p.type,
+        });
+        break;
+      }
+    }
+  }
+
+  // find excerpt range: focus on LAST ##[error] line
+  let errorLine = -1;
+  for (let i = lines.length - 1; i >= 0; i--) {
+    if (/##\[error\]/i.test(lines[i])) {
+      errorLine = i;
+      break;
+    }
+  }
+
+  let start: number;
+  let end: number;
+
+  if (errorLine === -1) {
+    start = Math.max(0, totalLines - excerptLines);
+    end = totalLines;
+  } else {
+    const contextAfter = 5;
+    const contextBefore = excerptLines - contextAfter;
+    start = Math.max(0, errorLine - contextBefore);
+    end = Math.min(totalLines, errorLine + contextAfter);
+  }
+
+  return {
+    totalLines,
+    index,
+    excerpt: {
+      content: lines.slice(start, end).join("\n"),
+      startLine: start + 1,
+      endLine: end,
+    },
+  };
+}
+
+type JobLogResult = {
+  job_id: number;
+  job_name: string;
+  job_url: string;
+  failed_steps: string[];
+  log_index: LogLine[];
+  excerpt: {
+    start_line: number;
+    end_line: number;
+    total_lines: number;
+    content: string;
+  };
+  full_log_path: string;
+};
+
 export function GetCheckSuiteLogsTool(ctx: ToolContext) {
  return tool({
    name: "get_check_suite_logs",
    description:
-      "get workflow run logs for a failed check suite. pass check_suite.id from the webhook payload.",
+      "get workflow run logs for a failed check suite. returns a log_index of interesting lines, " +
+      "a curated excerpt, and full_log_path for deeper investigation. " +
+      "pass check_suite.id from the webhook payload.",
    parameters: GetCheckSuiteLogs,
-    execute: execute(async ({ check_suite_id }) => {
+    execute: execute(async (params) => {
+      const check_suite_id = params.check_suite_id;
+
      // get workflow runs for this specific check suite
      const workflowRuns = await ctx.octokit.paginate(
        ctx.octokit.rest.actions.listWorkflowRunsForRepo,
@@ -30,67 +147,101 @@ export function GetCheckSuiteLogsTool(ctx: ToolContext) {
        return {
          check_suite_id,
          message: "no failed workflow runs found for this check suite",
-          workflow_runs: [],
+          failed_jobs: [],
        };
      }

+      // setup logs directory
+      const tempDir = process.env.PULLFROG_TEMP_DIR;
+      if (!tempDir) {
+        throw new Error("PULLFROG_TEMP_DIR not set");
+      }
+      const logsDir = join(tempDir, "ci-logs");
+      mkdirSync(logsDir, { recursive: true });
+
+      const jobResults: JobLogResult[] = [];
+
      // get logs for each failed run
-      const logsForRuns = await Promise.all(
-        failedRuns.map(async (run) => {
-          const jobs = await ctx.octokit.paginate(ctx.octokit.rest.actions.listJobsForWorkflowRun, {
-            owner: ctx.repo.owner,
-            repo: ctx.repo.name,
-            run_id: run.id,
-          });
+      for (const run of failedRuns) {
+        const jobs = await ctx.octokit.paginate(ctx.octokit.rest.actions.listJobsForWorkflowRun, {
+          owner: ctx.repo.owner,
+          repo: ctx.repo.name,
+          run_id: run.id,
+        });

-          const jobLogs = await Promise.all(
-            jobs.map(async (job) => {
-              try {
-                const logsResponse = await ctx.octokit.rest.actions.downloadJobLogsForWorkflowRun({
-                  owner: ctx.repo.owner,
-                  repo: ctx.repo.name,
-                  job_id: job.id,
-                });
+        // only process failed jobs
+        const failedJobs = jobs.filter((job) => job.conclusion === "failure");

-                const logsUrl = logsResponse.url;
-                const logsText = await fetch(logsUrl).then((r) => r.text());
+        for (const job of failedJobs) {
+          try {
+            const logsResponse = await ctx.octokit.rest.actions.downloadJobLogsForWorkflowRun({
+              owner: ctx.repo.owner,
+              repo: ctx.repo.name,
+              job_id: job.id,
+            });

-                return {
-                  job_id: job.id,
-                  job_name: job.name,
-                  status: job.status,
-                  conclusion: job.conclusion,
-                  started_at: job.started_at,
-                  completed_at: job.completed_at,
-                  logs: logsText,
-                };
-              } catch (error) {
-                return {
-                  job_id: job.id,
-                  job_name: job.name,
-                  status: job.status,
-                  conclusion: job.conclusion,
-                  started_at: job.started_at,
-                  completed_at: job.completed_at,
-                  error: `failed to fetch logs: ${error}`,
-                };
-              }
-            })
-          );
+            const logsUrl = logsResponse.url;
+            const logsText = await fetch(logsUrl).then((r) => r.text());

-          return {
-            workflow_run_id: run.id,
-            workflow_name: run.name,
-            html_url: run.html_url,
-            conclusion: run.conclusion,
-            jobs: jobLogs,
-          };
-        })
-      );
+            // write full log to disk
+            const logPath = join(logsDir, `job-${job.id}.log`);
+            writeFileSync(logPath, logsText);
+
+            // analyze log
+            const analysis = analyzeLog(logsText, 80);
+
+            // get failed steps
+            const failedSteps =
+              job.steps
+                ?.filter((s) => s.conclusion === "failure")
+                .map((s) => `Step ${s.number}: ${s.name}`) ?? [];
+
+            jobResults.push({
+              job_id: job.id,
+              job_name: job.name,
+              job_url: job.html_url ?? "",
+              failed_steps: failedSteps,
+              log_index: analysis.index,
+              excerpt: {
+                start_line: analysis.excerpt.startLine,
+                end_line: analysis.excerpt.endLine,
+                total_lines: analysis.totalLines,
+                content: analysis.excerpt.content,
+              },
+              full_log_path: logPath,
+            });
+
+            log.debug(`analyzed logs for job ${job.name}: ${analysis.index.length} indexed lines`);
+          } catch (error) {
+            log.error(`failed to fetch logs for job ${job.id}: ${error}`);
+          }
+        }
+      }

      return {
+        _instructions: {
+          overview:
+            "this result contains CI failure information. use log_index to find interesting lines, then read full_log_path for details.",
+          fields: {
+            log_index:
+              "array of interesting lines (errors, warnings, failures) with line numbers. use these to navigate the full log.",
+            excerpt:
+              "a curated ~80 line window around the last error. may not show all failures if they occur in different places.",
+            full_log_path:
+              "path to the complete log file. read specific line ranges using the line numbers from log_index.",
+            failed_steps:
+              "which CI steps failed. read the workflow yml to understand what commands these steps run.",
+          },
+          workflow: [
+            "1. scan log_index to see where errors/warnings/failures are located",
+            "2. read excerpt for immediate context",
+            "3. if excerpt doesn't show what you need, read specific line ranges from full_log_path",
+            "4. check failed_steps to understand what command failed",
+          ],
+        },
        check_suite_id,
-        workflow_runs: logsForRuns,
+        repo: `${ctx.repo.owner}/${ctx.repo.name}`,
+        failed_jobs: jobResults,
      };
    }),
  });