fix(server): generate personas through the built-in proxy (UF-045)

POST /api/personas/generate posted to /api/anthropic/v1/messages, a route
that does not exist, with no session token and a port built from
address().toString(), so the URL did not even parse. It also pinned
claude-sonnet-4-6, which needs an Anthropic key.

The route now calls the built-in proxy through self-request.ts, reads the
OpenAI-format reply, and picks the model from the user's default with the
same fallback chat uses, so a local Ollama model works too. An end-to-end
test with localhost trust off checks the generated persona comes back.
This commit is contained in:
shumeiko.it
2026-10-26 01:50:47 +01:00
parent d6217ad7c2
commit 48bdab8087
3 changed files with 96 additions and 12 deletions

View File

@@ -73,7 +73,7 @@ Baseline: commit `b20b138` of `git.os-dev.tech/oleg.m/waggle-os`.
| UF-042 | Sidebar hints repeat the visible labels and are partly hidden behind the main area | Fixed | Web UI |
| UF-043 | The chat composer draws two focus outlines | Fixed | Web UI |
| UF-044 | A working API key shows as not responding, and other internal model calls fail | Fixed | Sidecar, models, auth |
| UF-045 | Generating a persona from a description calls a route that does not exist | Open | Sidecar, agents |
| UF-045 | Generating a persona from a description calls a route that does not exist | Fixed | Sidecar, agents |
## Entries
@@ -448,10 +448,10 @@ Entries UF-010 to UF-019 come from the UI audit `anti-slop/audit-001-2026-10-14.
- Recommendation: apply the same change in the main project, and run the auth-sensitive suites with localhost trust off.
### UF-045. Generating a persona from a description calls a route that does not exist
- Status: Open
- Status: Fixed
- Found: 2026-10-26
- Where: at commit `b20b138`, `packages/server/src/local/routes/personas.ts:160` (`POST /api/personas/generate`, used by the agent templates view)
- Symptom: generating an agent persona from a text description cannot get an answer from the model.
- Cause: the route posts to `/api/anthropic/v1/messages`, which no route serves. The request also has no session token, and its port comes from `address().toString()`, which is `[object Object]`.
- What we did: nothing yet. The fix means moving the call to the built-in proxy (`/v1/chat/completions` through `self-request.ts`) and reading its OpenAI-format reply.
- Recommendation: move the call to the built-in proxy in the main project as well.
- Cause: the route posts to `/api/anthropic/v1/messages`, which no route serves. The request also has no session token, and its port comes from `address().toString()`, which is `[object Object]`, so the URL does not even parse. The model was fixed to `claude-sonnet-4-6`, so without an Anthropic key the call could not work either.
- What we did: the route calls the built-in proxy (`/v1/chat/completions` through `self-request.ts`, UF-044) and reads its OpenAI-format reply with `parseOpenAiTextCompletion`. The model starts from the user's default and falls back through `resolveUsableModel`, the way chat does, so a local Ollama model works too. An end-to-end test with localhost trust off checks that the persona the model wrote comes back; before the fix it got HTTP 500.
- Recommendation: apply the same change in the main project.

View File

@@ -1,14 +1,18 @@
import type { FastifyPluginAsync } from 'fastify';
import { z } from 'zod';
import { WaggleConfig } from '@waggle/core';
import {
listPersonas,
getPersona,
saveCustomPersona,
deleteCustomPersona,
isValidCustomPersonaId,
parseOpenAiTextCompletion,
type AgentPersona,
} from '@waggle/agent';
import { validateBody } from '../../validate-body.js';
import { resolveUsableModel } from '../model-availability.js';
import { selfBaseUrl, selfRequestHeaders } from '../self-request.js';
const customPersonaIdSchema = z.string().max(200).refine(isValidCustomPersonaId, {
message: 'Invalid custom persona ID',
@@ -156,13 +160,17 @@ User's description: "${body.prompt}"
Respond with ONLY valid JSON, no markdown or explanation.`;
// Use the built-in Anthropic proxy or LiteLLM
const chatUrl = `http://localhost:${fastify.server.address()?.toString().split(':').pop() ?? '3333'}/api/anthropic/v1/messages`;
const res = await fetch(chatUrl, {
// The built-in proxy routes every configured provider, Ollama included.
// Start from the user's default model and fall back the way chat does.
const model = await resolveUsableModel(
fastify,
new WaggleConfig(fastify.localConfig.dataDir).getDefaultModel() || 'claude-sonnet-4-6',
);
const res = await fetch(`${selfBaseUrl(fastify)}/v1/chat/completions`, {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
headers: selfRequestHeaders(fastify),
body: JSON.stringify({
model: 'claude-sonnet-4-6',
model,
max_tokens: 2000,
messages: [{ role: 'user', content: metaPrompt }],
}),
@@ -172,8 +180,7 @@ Respond with ONLY valid JSON, no markdown or explanation.`;
return reply.code(502).send({ error: 'LLM request failed' });
}
const result = await res.json() as { content?: Array<{ text?: string }> };
const text = result.content?.[0]?.text ?? '';
const text = parseOpenAiTextCompletion(await res.json()).content;
// Extract JSON from response
const jsonMatch = text.match(/\{[\s\S]*\}/);

View File

@@ -0,0 +1,77 @@
/**
* UF-045 — generating a persona from a description goes through the built-in
* proxy with the session token, and reads the proxy's OpenAI-format reply.
* Runs against a listening server with localhost trust off; only the upstream
* Anthropic call is faked.
*/
import { describe, it, expect, beforeAll, afterAll, afterEach, vi } from 'vitest';
import fs from 'node:fs';
import os from 'node:os';
import path from 'node:path';
import type { FastifyInstance } from 'fastify';
import { buildLocalServer } from '../../src/local/index.js';
import { injectWithAuth } from '../test-utils.js';
describe('POST /api/personas/generate (UF-045)', () => {
let server: FastifyInstance;
let tmpDir: string;
const realFetch = globalThis.fetch;
const priorTrustLocalhost = process.env.WAGGLE_TRUST_LOCALHOST;
beforeAll(async () => {
process.env.WAGGLE_TRUST_LOCALHOST = '0';
tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'waggle-personas-gen-'));
server = await buildLocalServer({ dataDir: tmpDir });
await server.listen({ port: 0, host: '127.0.0.1' });
server.vault.set('anthropic', 'sk-ant-test-key');
// startService sets the provider status; buildLocalServer alone does not.
server.agentState.llmProvider = {
provider: 'anthropic-proxy',
health: 'degraded',
detail: 'Built-in provider proxy (verification pending)',
checkedAt: new Date().toISOString(),
};
});
afterEach(() => vi.unstubAllGlobals());
afterAll(async () => {
await server.close();
fs.rmSync(tmpDir, { recursive: true, force: true });
if (priorTrustLocalhost === undefined) delete process.env.WAGGLE_TRUST_LOCALHOST;
else process.env.WAGGLE_TRUST_LOCALHOST = priorTrustLocalhost;
});
it('returns the persona the model wrote', async () => {
const persona = {
name: 'Release Scout',
description: 'Tracks what changed between releases.',
icon: '🔭',
systemPrompt: 'You compare release notes and summarise breaking changes.',
tools: ['web_search', 'document_processing'],
};
const upstream = vi.fn(async () => new Response(JSON.stringify({
id: 'msg_persona',
type: 'message',
role: 'assistant',
model: 'claude-sonnet-4-6',
content: [{ type: 'text', text: JSON.stringify(persona) }],
stop_reason: 'end_turn',
usage: { input_tokens: 10, output_tokens: 40 },
}), { status: 200, headers: { 'content-type': 'application/json' } }));
vi.stubGlobal('fetch', (input: RequestInfo | URL, init?: RequestInit) => {
const url = typeof input === 'string' ? input : input instanceof URL ? input.href : input.url;
return url.startsWith('https://api.anthropic.com/') ? upstream() : realFetch(input, init);
});
const res = await injectWithAuth(server, {
method: 'POST',
url: '/api/personas/generate',
payload: { prompt: 'An agent that watches our release notes' },
});
expect(res.statusCode).toBe(200);
expect(JSON.parse(res.body)).toEqual(persona);
expect(upstream).toHaveBeenCalledTimes(1);
});
});