next-issue 2.5.1 → 2.5.2

package/README.md

@@ -64,8 +64,9 @@ and exits with code 1. For each issue:
    reviewer reads files and runs commands in the worktree, so it can prove a
    finding with the tests or the type check. The harness then throws away every
    change that the reviewer left, so no such change can reach a later commit.
-   The reviewer returns its verdict with a `submit_result` tool: a summary and one entry per
-   finding, each rated `blocking` or `minor`. The findings are the only ground
+   The reviewer ends its last message with its verdict as a json block: a
+   summary and one entry per finding, each rated `blocking` or `minor`. A
+   reply without a valid block hands the issue to a person. The findings are the only ground
    truth. The reviewer gives no approve or reject field, because such a field
    can contradict the findings. The harness puts the result on the pull
    request.

package/dist/agents.mjs

@@ -1,5 +1,5 @@
 import { join, sep } from 'node:path';
-import { createAgentSession, DefaultResourceLoader, defineTool, getAgentDir, ModelRuntime, SessionManager, } from '@earendil-works/pi-coding-agent';
+import { createAgentSession, DefaultResourceLoader, getAgentDir, ModelRuntime, SessionManager, } from '@earendil-works/pi-coding-agent';
 import { onStop } from './exec.mjs';
 import { emptyUsage } from './observe.mjs';
 const REVIEWING = {
@@ -9,7 +9,6 @@ const CODING = {
     tools: ['+grep', '+find', '+ls'],
 };
 export { CODING, REVIEWING };
-export const SUBMIT_TOOL = 'submit_result';
 export function isOwnExtension(path) {
     return path.split(sep).includes('pi-next-issue');
 }
@@ -19,24 +18,6 @@ export async function runAgent(recorder, request) {
     const agentDir = runtime.agentDir ?? getAgentDir();
     const started = Date.now();
     log.event('agent.start', { model: request.model, timeoutMs: request.timeoutMs }, 'normal');
-    let structured;
-    const customTools = [];
-    if (request.outputSchema !== undefined) {
-        customTools.push(defineTool({
-            name: SUBMIT_TOOL,
-            label: 'Submit result',
-            description: 'Give the result. Call it once, as your last action.',
-            parameters: request.outputSchema,
-            async execute(_toolCallId, params) {
-                structured = params;
-                return {
-                    content: [{ type: 'text', text: 'Result saved.' }],
-                    details: {},
-                    terminate: true,
-                };
-            },
-        }));
-    }
     const resourceLoader = new DefaultResourceLoader({
         cwd: request.cwd,
         agentDir,
@@ -58,7 +39,6 @@ export async function runAgent(recorder, request) {
         resourceLoader,
         sessionManager: SessionManager.create(request.cwd, recorder.sessionDir),
         tools: request.profile.tools,
-        customTools,
         modelRuntime,
         ...(runtime.settingsManager === undefined ? {} : { settingsManager: runtime.settingsManager }),
         ...(model === undefined ? {} : { model }),
@@ -105,7 +85,6 @@ export async function runAgent(recorder, request) {
             costUsd: stats.cost,
             model: name,
             sessionId: stats.sessionId,
-            structured,
         };
     }
     finally {

package/dist/pipeline.mjs

@@ -7,7 +7,7 @@ import { message } from './observe.mjs';
 import { capDiff, fixPrompt, hookPrompt, implementPrompt, parseCommitSubject, parseUnrelated, reviewPrompt, } from './prompts.mjs';
 import { detectSetupCommand } from './setup.mjs';
 import { dropState, readState, resetState, savedIssues, writeState, } from './state.mjs';
-import { blockingFindings, fingerprint, formatFindings, formatVerdict, openFindings, readVerdict, VERDICT_SCHEMA, } from './verdict.mjs';
+import { blockingFindings, fingerprint, formatFindings, formatVerdict, openFindings, parseVerdict, } from './verdict.mjs';
 export const PORTS = {
     addWorktree,
     commentOnIssue,
@@ -362,12 +362,11 @@ async function reviewRound(job) {
         profile: REVIEWING,
         model: config.models.reviewer,
         timeoutMs: config.agentTimeoutMinutes * 60_000,
-        outputSchema: VERDICT_SCHEMA,
     }));
     if (await ports.discardChanges(worktree, before)) {
         log.event('review.discard', { round: state.reviewRounds }, 'quiet');
     }
-    const verdict = readVerdict(review.structured);
+    const verdict = parseVerdict(review.text);
     if (verdict === undefined) {
         await escalate(job, 'The reviewer gave no verdict.');
         return { done: true, report: await finish(job, 'needs-human', 'no verdict') };

package/dist/prompts.mjs

@@ -60,7 +60,8 @@ export function reviewPrompt(issue, comments, diff, earlier) {
         'The harness discards each change that you make, so a change is only lost work.',
         scope,
         'Mark a finding blocking only for a wrong result, a missing part of the issue, a regression or a broken convention.',
-        'Give the result with the submit_result tool: a short summary and one entry per finding. The findings decide the outcome, so a problem that you leave out counts as no problem.',
+        'End your final message with the result as a json block in a ```json fence, in the form {"summary": "<short summary>", "findings": [{"severity": "blocking" or "minor", "detail": "<what is wrong and where>"}]}.',
+        'Give one entry per finding, and an empty list when nothing is wrong. The findings decide the outcome, so a problem that you leave out counts as no problem.',
     ].join('\n\n');
 }
 const UNRELATED_RULE = [

package/dist/verdict.mjs

@@ -1,26 +1,3 @@
-export const VERDICT_SCHEMA = {
-    type: 'object',
-    properties: {
-        summary: { type: 'string', description: 'Short summary of the review' },
-        findings: {
-            type: 'array',
-            description: 'One entry per problem. Empty when nothing is wrong.',
-            items: {
-                type: 'object',
-                properties: {
-                    severity: {
-                        type: 'string',
-                        enum: ['blocking', 'minor'],
-                        description: 'blocking for a wrong result, a missing part of the issue, a regression or a broken convention; minor for taste and style',
-                    },
-                    detail: { type: 'string', description: 'What is wrong and where' },
-                },
-                required: ['severity', 'detail'],
-            },
-        },
-    },
-    required: ['summary', 'findings'],
-};
 const SEVERITIES = ['blocking', 'minor'];
 export function readVerdict(value) {
     if (typeof value !== 'object' || value === null) {
@@ -45,6 +22,19 @@ export function readVerdict(value) {
     }
     return { summary, findings: checked };
 }
+const JSON_BLOCK = /```json[^\n]*\n([\s\S]*?)```/g;
+export function parseVerdict(text) {
+    const last = [...text.matchAll(JSON_BLOCK)].at(-1);
+    if (last === undefined) {
+        return undefined;
+    }
+    try {
+        return readVerdict(JSON.parse(last[1]));
+    }
+    catch {
+        return undefined;
+    }
+}
 export function blockingFindings(verdict) {
     return verdict.findings.filter((finding) => finding.severity === 'blocking');
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "next-issue",
-  "version": "2.5.1",
+  "version": "2.5.2",
   "description": "Autonomous issue harness driven by the pi SDK",
   "keywords": [
     "github",