claude
f0173440e4
Piggy was a hand-rolled OpenAI tool loop. It is now a Prime Agent session — Prime Intellect's own harness, embedded as a Node library — answering from PIG's tools and, for the first time, able to put information into the CRM rather than only read it out. The harness is a coding agent, so the first job was taking the coding agent away from it. `noTools: 'all'` plus an explicit allowlist leaves the model with PIG's ten `pig_*` tools and no bash, no filesystem, no IPython. That holds under attack: a hostile extension, a skill and a settings file planted in the agent's own directory, then `setActiveToolsByName` called with every built-in, still leaves ten tools, all ours. Both lines are load-bearing — `noTools` alone registers nothing, and the allowlist is what admits our own. Writing is gated rather than assumed. A change is proposed, not made: the tool returns a description, the transcript renders a diff card, and nothing reaches the database until someone presses Apply. Contracts, commitments, allocations and compliance always stop for a human whatever the mode. Every write runs through `executeMutation` as the calling user, so their capabilities and the audit trail apply exactly as they would to a human's. Four things about the SDK are wrong in its own documentation and cost a debugging cycle each: models.json does not resolve an env var name for `apiKey`, it sends the literal string; there is no built-in prime-inference provider in 0.84.1; a ResourceLoader you pass in is never reloaded for you; and the stock system prompt is a coding-assistant prompt that must be replaced — but replacing it also silently removes the tool list, because the harness only renders that section when it owns the prompt. AGENTS.md records all four. The expensive one was thinking level. The harness defaults to `medium`, and nemotron spent an entire 4,096-token budget reasoning and returned an empty answer. `low` was worse; `off` omits the parameter so the endpoint's default wins. An explicit `reasoning_effort: none` via `thinkingLevelMap` took a turn from 6,195 output tokens to 149. And a turn is now bounded. The harness loop is `while (true)` with no iteration cap; a runaway on a frontier model would have eaten the credit it is supposed to report on. Ceilings on model calls and tokens, enforced both through the harness hook and independently from the event stream, plus a per-user daily spend limit — and the ledger now records spend on turns that fail, which it previously discarded. Signing in lands on /piggy, which is a workspace: conversations down one side, the agent in the middle, what it did and what it cost beside it. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
47 lines
2.3 KiB
TypeScript
47 lines
2.3 KiB
TypeScript
import assert from 'node:assert/strict';
|
|
import test from 'node:test';
|
|
import { buildPiggySystemPrompt } from '../src/agent/prompt';
|
|
import { assertPigToolBoundary } from '../src/chat';
|
|
|
|
/**
|
|
* What is left of this file after the harness swap.
|
|
*
|
|
* The hand-rolled loop that used to be tested here — the SSE reader, the
|
|
* tool-call assembler, the retry budget — belongs to Prime Agent now, and its
|
|
* tests went with it. Two things did not move, and both are the sort that fail
|
|
* silently rather than loudly.
|
|
*/
|
|
|
|
test('ambient coding tools are rejected at the boundary', () => {
|
|
assert.throws(
|
|
() => assertPigToolBoundary([{ name: 'pig_get_idle_capacity' }, { name: 'bash' }]),
|
|
/outside the PIG tool boundary/,
|
|
);
|
|
// A tool that starts pig_ but reads like a filesystem is refused too: the
|
|
// prefix is a convention, and a convention alone is not a boundary.
|
|
assert.throws(() => assertPigToolBoundary([{ name: 'pig_file_write' }]), /outside the PIG tool boundary/);
|
|
assert.throws(() => assertPigToolBoundary([{ name: 'pig_shell_exec' }]), /outside the PIG tool boundary/);
|
|
assert.doesNotThrow(() =>
|
|
assertPigToolBoundary([{ name: 'pig_get_idle_capacity' }, { name: 'pig_log_activity' }]),
|
|
);
|
|
});
|
|
|
|
test('the prompt Piggy actually runs on still states the units rule and the margin definitions', () => {
|
|
const prompt = buildPiggySystemPrompt({ mode: 'read_only' });
|
|
|
|
// The whole point: 189 spoken as "$189 per GPU-hour" is a hundredfold error
|
|
// on the number everyone in the room is watching. This assertion survived the
|
|
// move from the retired chat loop to `agent/prompt.ts` because the failure it
|
|
// guards against did not.
|
|
assert.match(prompt, /ends in Cents is an integer number of US cents/i);
|
|
assert.match(prompt, /costPerGpuHourCents: 189 is \$1\.89 per GPU-hour/);
|
|
assert.match(prompt, /ends in Pct, and utilisation, is a share between 0 and 1/);
|
|
// Margin against sold hours only would report a losing block as healthy.
|
|
assert.match(prompt, /revenue minus the FULL cost of the commitment/);
|
|
assert.match(prompt, /REMAINING unsold hours must fetch/);
|
|
// And the stock harness preamble, which introduces a coding assistant with a
|
|
// filesystem, must be gone rather than merely appended to.
|
|
assert.match(prompt, /no shell, filesystem, browser, code execution, or hidden tools/i);
|
|
assert.doesNotMatch(prompt, /coding assistant/i);
|
|
});
|