profit/lakehouse
1
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases Wiki Activity

remove 5 orphaned dev experiments from tests/real-world/

Browse Source 
Per J: "all of our test code ended up in the main." These are 5 one-time
dev experiments that were never wired into any automation and have zero
live consumers in the production code path. Deleting them.

Removed (1418 LOC total):
- enrich_prd_pipeline.ts      (528 LOC) — Phase 21 architecture stress test
- nine_consecutive_audits.ts  (185 LOC) — empirical study of audit compounding
- hard_task_escalation.ts     (267 LOC) — escalation-ladder test (refs retired
                                            cloud models qpt-oss:20b/120b)
- autonomous_loop.ts          (214 LOC) — wrapper experiment around scrum_master
- consensus_reducer_design.ts (224 LOC) — N=3 design consultation; output JSON
                                            referenced from pathway_memory.rs
                                            comment but the script itself has
                                            no consumer

Verified: 0 references in package.json / justfile / Makefile / any
production .ts/.rs/.sh file. The single mention from pathway_memory.rs
is a //! doc comment referencing the JSON output (data/_kb/
consensus_reducer_design_*.json), not the script. Build clean post-delete.

KEPT:
- scrum_master_pipeline.ts — referenced from observer.ts, vectord, scripts
- scrum_applier.ts          — referenced from auditor schemas

If you need any of these back, they're in git history. cherry-pick or
git show HEAD~1 -- tests/real-world/<file>.ts will recover the source.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
root
2026-05-03 02:05:24 -05:00
co-authored by Claude Opus 4.7
parent bb5a3b3f5e
commit 6aafd41785
5 changed files with 0 additions and 1419 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/real-world/autonomous_loop.ts
-214
View File
@@ -1,214 +0,0 @@
#!/usr/bin/env bun
// Autonomous scrum loop — wraps scrum_master_pipeline.ts + scrum_applier.ts
// in a goal-driven retry loop. Observer is POSTed an iteration summary at
// every boundary so it can build meta-commentary outside the loop's epistemic
// scope.
//
// Usage:
// LOOP_TARGETS="crates/a/src/x.rs,crates/b/src/y.rs" \
// LOOP_MAX_ITERS=5 \
// LOOP_PUSH=1 \
// bun run tests/real-world/autonomous_loop.ts
//
// Stop conditions: max_iters reached OR 2 consecutive iters with 0 commits.
import { spawn } from "node:child_process";
import { appendFile, readFile } from "node:fs/promises";
import { existsSync } from "node:fs";
const REPO = "/home/profit/lakehouse";
const OBSERVER = process.env.LOOP_OBSERVER ?? "http://localhost:3800";
const BRANCH = process.env.LOOP_BRANCH ?? "scrum/auto-apply-19814";
const MAX_ITERS = Number(process.env.LOOP_MAX_ITERS ?? 3);
const PUSH = process.env.LOOP_PUSH === "1";
const MIN_CONF = process.env.LOOP_MIN_CONF ?? "85";
// Optional override — when unset, let scrum_applier.ts use ITS default
// (currently x-ai/grok-4.1-fast on openrouter). The prior hardcoded
// qwen3-coder:480b default was clobbering the applier's own default
// and forcing every iter to hit the throttled ollama_cloud account.
const APPLIER_MODEL = process.env.LOOP_APPLIER_MODEL;
const APPLIER_PROVIDER = process.env.LOOP_APPLIER_PROVIDER;
const TARGETS = (process.env.LOOP_TARGETS ?? "crates/queryd/src/service.rs,crates/gateway/src/main.rs,crates/gateway/src/v1/mod.rs")
.split(",").map(s => s.trim()).filter(Boolean);
const FORENSIC = process.env.LH_SCRUM_FORENSIC ?? `${REPO}/docs/SCRUM_FORENSIC_PROMPT.md`;
const PROPOSAL = process.env.LH_SCRUM_PROPOSAL ?? `${REPO}/docs/SCRUM_FIX_WAVE.md`;
const LOOP_ID = `loop_${Date.now().toString(36)}`;
const JOURNAL = `${REPO}/data/_kb/autonomous_loops.jsonl`;
interface IterResult {
iter: number;
scrum_reviews_added: number;
applier_outcomes: Record<string, number>;
commits_landed: number;
commit_shas: string[];
build_status: "green" | "red" | "unknown";
duration_ms: number;
}
function log(msg: string) {
const ts = new Date().toISOString().slice(11, 19);
console.log(`[loop ${LOOP_ID} ${ts}] ${msg}`);
}
function runCmd(cmd: string, args: string[], env: Record<string, string> = {}): Promise<{ code: number; stdout: string; stderr: string }> {
return new Promise((resolve) => {
const child = spawn(cmd, args, { cwd: REPO, env: { ...process.env, ...env } });
let stdout = "", stderr = "";
child.stdout.on("data", (d) => { stdout += d; process.stdout.write(d); });
child.stderr.on("data", (d) => { stderr += d; process.stderr.write(d); });
child.on("close", (code) => resolve({ code: code ?? -1, stdout, stderr }));
});
}
async function countLines(path: string): Promise<number> {
if (!existsSync(path)) return 0;
const text = await readFile(path, "utf8");
return text.split("\n").filter(Boolean).length;
}
async function gitHeadSha(): Promise<string> {
const r = await runCmd("git", ["rev-parse", "HEAD"]);
return r.stdout.trim();
}
async function commitsSince(baseSha: string): Promise<string[]> {
const r = await runCmd("git", ["log", "--oneline", `${baseSha}..HEAD`]);
return r.stdout.trim().split("\n").filter(Boolean);
}
async function cargoCheckGreen(): Promise<boolean> {
log("cargo check --workspace …");
const r = await runCmd("cargo", ["check", "--workspace", "--quiet"]);
return r.code === 0;
}
async function postObserver(payload: object) {
try {
const r = await fetch(`${OBSERVER}/event`, {
method: "POST",
headers: { "Content-Type": "application/json" },
body: JSON.stringify(payload),
signal: AbortSignal.timeout(5000),
});
if (!r.ok) log(`observer POST returned ${r.status}`);
} catch (e: any) {
log(`observer POST failed: ${e.message}`);
}
}
async function runIter(iter: number, baseSha: string): Promise<IterResult> {
const t0 = Date.now();
log(`══ iter ${iter} start (base ${baseSha.slice(0, 8)}) targets=${TARGETS.length}`);
const reviewsBefore = await countLines(`${REPO}/data/_kb/scrum_reviews.jsonl`);
const applyBefore = await countLines(`${REPO}/data/_kb/auto_apply.jsonl`);
log(`scrum_master_pipeline.ts → ${TARGETS.length} files`);
await runCmd("bun", ["run", "tests/real-world/scrum_master_pipeline.ts"], {
LH_SCRUM_FILES: TARGETS.join(","),
LH_SCRUM_FORENSIC: FORENSIC,
LH_SCRUM_PROPOSAL: PROPOSAL,
});
log(`scrum_applier.ts COMMIT=1 MIN_CONF=${MIN_CONF} files=${TARGETS.length}`);
// Only forward model/provider when explicitly overridden — otherwise
// let scrum_applier.ts use its own defaults (Grok 4.1 fast on openrouter).
const applierEnv: Record<string, string> = {
LH_APPLIER_COMMIT: "1",
LH_APPLIER_MIN_CONF: MIN_CONF,
LH_APPLIER_MAX_FILES: String(TARGETS.length),
LH_APPLIER_BRANCH: BRANCH,
// Constrain applier to THIS iter's targets so it patches what we
// just reviewed instead of the highest-confidence file from history.
LH_APPLIER_FILES: TARGETS.join(","),
};
if (APPLIER_MODEL) applierEnv.LH_APPLIER_MODEL = APPLIER_MODEL;
if (APPLIER_PROVIDER) applierEnv.LH_APPLIER_PROVIDER = APPLIER_PROVIDER;
await runCmd("bun", ["run", "tests/real-world/scrum_applier.ts"], applierEnv);
const reviewsAfter = await countLines(`${REPO}/data/_kb/scrum_reviews.jsonl`);
const applyAfterText = existsSync(`${REPO}/data/_kb/auto_apply.jsonl`)
? await readFile(`${REPO}/data/_kb/auto_apply.jsonl`, "utf8")
: "";
const applyRows = applyAfterText.split("\n").filter(Boolean).slice(applyBefore);
const outcomes: Record<string, number> = {};
for (const line of applyRows) {
try {
const o = JSON.parse(line);
outcomes[o.action ?? "?"] = (outcomes[o.action ?? "?"] ?? 0) + 1;
} catch { /* skip malformed */ }
}
const commitShas = await commitsSince(baseSha);
const buildStatus = commitShas.length > 0 ? (await cargoCheckGreen() ? "green" : "red") : "unknown";
const result: IterResult = {
iter,
scrum_reviews_added: reviewsAfter - reviewsBefore,
applier_outcomes: outcomes,
commits_landed: commitShas.length,
commit_shas: commitShas.map(s => s.split(" ")[0]),
build_status: buildStatus,
duration_ms: Date.now() - t0,
};
log(`iter ${iter} done — reviews+${result.scrum_reviews_added} commits=${result.commits_landed} build=${buildStatus} (${(result.duration_ms / 1000).toFixed(1)}s)`);
await postObserver({
source: "autonomous_loop",
loop_id: LOOP_ID,
event_kind: "iteration_complete",
iter,
targets: TARGETS,
success: buildStatus !== "red",
scrum_reviews_added: result.scrum_reviews_added,
applier_outcomes: result.applier_outcomes,
commits_landed: result.commits_landed,
commit_shas: result.commit_shas,
build_status: buildStatus,
duration_ms: result.duration_ms,
ts: new Date().toISOString(),
});
await appendFile(JOURNAL, JSON.stringify({ loop_id: LOOP_ID, ...result, ts: new Date().toISOString() }) + "\n");
return result;
}
async function main() {
log(`autonomous loop starting · branch=${BRANCH} max_iters=${MAX_ITERS} push=${PUSH}`);
log(`targets: ${TARGETS.join(", ")}`);
const branchR = await runCmd("git", ["branch", "--show-current"]);
if (branchR.stdout.trim() !== BRANCH) {
log(`ERROR: on branch ${branchR.stdout.trim()}, expected ${BRANCH}. Refusing to run.`);
process.exit(1);
}
let consecutiveZero = 0;
for (let iter = 1; iter <= MAX_ITERS; iter++) {
const baseSha = await gitHeadSha();
const result = await runIter(iter, baseSha);
if (PUSH && result.commits_landed > 0) {
log(`git push origin ${BRANCH}`);
const pushR = await runCmd("git", ["push", "origin", BRANCH]);
if (pushR.code !== 0) log(`push failed (continuing): ${pushR.stderr.slice(0, 200)}`);
}
consecutiveZero = result.commits_landed === 0 ? consecutiveZero + 1 : 0;
if (consecutiveZero >= 2) {
log(`STOP: 2 consecutive iters with 0 commits. Loop converged or stuck.`);
break;
}
}
log(`loop ${LOOP_ID} complete. Journal: ${JOURNAL}`);
}
main().catch((e) => {
log(`FATAL: ${e.message}`);
process.exit(1);
});
tests/real-world/consensus_reducer_design.ts
-224
View File
@@ -1,224 +0,0 @@
// consensus_reducer_design.ts — N=3 design consultation.
//
// J's ask: enhance the tree-split reducer to preserve FULL backtrack-able
// context (endpoints tried, attempt count per model in the ladder, KB
// sources retrieved, context7 bridge hits, MCP observer signals, audit
// verdicts) instead of collapsing to a summary. Then index the full
// context through our existing vectord matrix indexing (HNSW + Lance +
// playbook_memory) so successful pathways become hot-swappable — the
// system asks "what did we try, what worked, in what order" for a
// similar task class and gets a ranked playbook back.
//
// Before building, consult three diverse models and print their proposals
// side-by-side so we can pick the convergent design.
const GATEWAY = "http://localhost:3100";
const DESIGN_BRIEF = `
# Context — Lakehouse signal→commit loop
We run 6x scrum-master iterations that audit Rust crates for PRD
alignment, produce findings + confidence, and feed an auto-applier that
lands small mechanical commits through a cargo-green-and-warning-stable
gate. Key components:
- \`tests/real-world/scrum_master_pipeline.ts\` — orchestrator. 9-rung
model LADDER (kimi-k2:1t → qwen3-coder:480b → deepseek → mistral-large
→ gpt-oss:120b → qwen3.5:397b → openrouter free rescues → local
qwen3.5:latest). Each target file retrieves 5 PRD chunks + 5
proposal-doc chunks via vectord RAG, tree-splits large files into 3.5K
shards, asks each rung in order, accepts first response passing
structural checks.
- \`mcp-server/observer.ts\` — receives scrum \`/event\` emissions
(file, verdict, critical_failures_count, gradient_tier, attempts,
reviewer_model, tree_split_fired). Escalates failure clusters to LLM
Team by POSTing to /v1/chat with qwen3-coder:480b.
- \`context7-bridge\` — external library docs lookup.
- \`auditor/audit.ts\` — independent N=3 consensus re-check of scrum
findings; writes to data/_kb/audit_facts.jsonl via LLM Team
\`/api/run?mode=extract\`.
- \`crates/vectord/src/playbook_memory.rs\` — indexing for proven
playbooks: PlaybookEntry, DocRef, FailureRecord, BoostEntry,
PatternReport. Uses HNSW index + Lance columnar backend + promotion
pipeline. Already battle-tested for workers/staffing queries.
- Tree-split REDUCER: after shards return map-style summaries, they are
concatenated with internal §N§ markers and fed to a reviewer model to
produce ONE file-level review. Currently the reducer sees summaries,
not the full context behind each shard's conclusion.
# The problem
The reducer currently TRUNCATES to a short summary. When the auditor or
a future iteration wants to backtrack WHY the reducer concluded what it
did — which attempt succeeded, which failed, what KB chunks were
retrieved, what observer signal classified the file as LOOPING vs
CONVERGING — that context is lost. So:
1. Auditor can't verify citation provenance beyond the summary line.
2. Applier can't tell a "tried X, failed, qwen fixed it" playbook from a
"tried X and it was easy" playbook — they look identical downstream.
3. The matrix indexing is only used for RAG chunks during the scrum
pass, NOT for storing the full end-to-end pathway of a successful
review.
# The design question
Propose an enhanced reducer + indexing design that:
(a) Preserves the FULL backtrack context per reviewed file:
- every ladder attempt (model, ms, accepted_y/n, reject_reason)
- every retrieved KB chunk (source doc, chunk id, cosine score, rank)
- every observer signal (class, priors, prior-iter outcomes)
- every context7 bridge hit (library, version pulled)
- every sub-pipeline call (LLM Team extract results, audit consensus)
(b) Stores this pathway into vectord's matrix indexing alongside the
review verdict so it becomes retrievable by similarity. When a new
file's fingerprint (task_class + file-path prefix + signal class)
matches a past successful pathway, the system can hot-swap by
replaying or short-circuiting to the model/KB combo that worked.
(c) Surfaces the matrix-index hit rate as a feedback signal on the
scrum's UI — "this file was solved 3 times before by the same ladder
rung; consider short-circuiting to rung 5."
(d) Is compatible with the existing playbook_memory.rs primitives
(PlaybookEntry, DocRef, FailureRecord, BoostEntry) — extend don't
replace. The indexing layer is in production for workers/staffing;
we want the reducer pathway to piggyback on proven infrastructure.
# Constraints
- NO new crate. Extend vectord + scrum_master_pipeline.
- Full context can be LARGE — a reviewed file might have 5 retrievals,
4 ladder attempts, 8 observer priors. Design the embedding /
fingerprint so similar-but-not-identical pathways cluster.
- The reducer summary is still needed for the reviewer LLM input —
don't remove it, ADD the full-context sidecar.
- Audit trail: every pathway must be replayable deterministically from
what's stored (i.e., enough context to re-run without the original
prompt cache).
# Required output (STRICT JSON, no prose, no markdown fences):
{
"approach": "one-paragraph summary of your proposed design",
"data_model": {
"new_fields_on_playbook_entry": [...],
"new_types": [ {"name": "...", "purpose": "...", "fields": [...]} ]
},
"storage_strategy": {
"what_to_vectorize": "the text that becomes the embedding",
"fingerprint_key": "the deterministic key for similarity retrieval",
"backend": "HNSW, Lance, playbook_memory — pick"
},
"reducer_changes": {
"inputs_added": [...],
"outputs_added": [...],
"compatibility_notes": "how existing callers stay working"
},
"hot_swap_logic": "concrete rule for when to skip the ladder and replay a past pathway",
"ui_signal": "what to surface so J sees whether matrix indexing is earning its keep",
"risks": [...],
"why_this_beats_summarization": "one-paragraph argument"
}
`.trim();
interface Probe {
name: string;
provider: "ollama" | "ollama_cloud" | "openrouter";
model: string;
}
// Round-3 probe set — 4 probes covering the remaining ladder rungs +
// architecture/provider diversity. J wanted all 4 of the untouched
// options so the aggregated 10-model signal is saturated across the
// usable ladder.
const PROBES: Probe[] = [
{ name: "qwen35-397b", provider: "ollama_cloud", model: "qwen3.5:397b" },
{ name: "openrouter-gpt-oss", provider: "openrouter", model: "openai/gpt-oss-120b:free" },
{ name: "openrouter-gemma3", provider: "openrouter", model: "google/gemma-3-27b-it:free" },
{ name: "qwen3-coder-480b-2", provider: "ollama_cloud", model: "qwen3-coder:480b" }, // second probe of the coding specialist — stability check
];
async function ask(p: Probe): Promise<{ name: string; raw: string; ms: number; error?: string }> {
const started = Date.now();
try {
const r = await fetch(`${GATEWAY}/v1/chat`, {
method: "POST",
headers: { "Content-Type": "application/json" },
body: JSON.stringify({
provider: p.provider,
model: p.model,
messages: [
{ role: "system", content: "You are a senior architect. Output STRICT JSON only." },
{ role: "user", content: DESIGN_BRIEF },
],
max_tokens: 3000,
temperature: 0,
}),
});
const ms = Date.now() - started;
if (!r.ok) return { name: p.name, raw: "", ms, error: `HTTP ${r.status}: ${(await r.text()).slice(0, 200)}` };
const j = await r.json();
const content = j.content ?? j.message?.content ?? j.choices?.[0]?.message?.content ?? "";
return { name: p.name, raw: String(content), ms };
} catch (e: any) {
return { name: p.name, raw: "", ms: Date.now() - started, error: String(e).slice(0, 200) };
}
}
function extractJson(raw: string): any | null {
let s = raw.trim();
const fence = s.match(/^```(?:json)?\s*/);
if (fence) s = s.slice(fence[0].length);
if (s.endsWith("```")) s = s.slice(0, -3).trim();
const first = s.indexOf("{");
const last = s.lastIndexOf("}");
if (first < 0 || last <= first) return null;
try {
return JSON.parse(s.slice(first, last + 1));
} catch {
return null;
}
}
function summarize(obj: any, max = 240): string {
if (!obj) return "(no JSON parsed)";
if (typeof obj === "string") return obj.length > max ? obj.slice(0, max) + "…" : obj;
if (Array.isArray(obj)) return obj.map((x) => summarize(x, max)).join("; ");
return Object.entries(obj)
.map(([k, v]) => `${k}=${summarize(v, max)}`)
.join(" | ");
}
async function main() {
console.log(`\n── N=3 design consensus ──`);
console.log(`models: ${PROBES.map((p) => p.model).join(", ")}\n`);
const results = await Promise.all(PROBES.map(ask));
for (const r of results) {
console.log(`\n── ${r.name} (${r.ms}ms) ──`);
if (r.error) { console.log(` ERROR: ${r.error}`); continue; }
const j = extractJson(r.raw);
if (!j) {
console.log(` raw (no JSON): ${r.raw.slice(0, 600)}…`);
continue;
}
console.log(` approach: ${summarize(j.approach, 400)}`);
console.log(` fingerprint: ${summarize(j.storage_strategy?.fingerprint_key, 200)}`);
console.log(` vectorize: ${summarize(j.storage_strategy?.what_to_vectorize, 200)}`);
console.log(` backend: ${summarize(j.storage_strategy?.backend, 200)}`);
console.log(` hot_swap: ${summarize(j.hot_swap_logic, 300)}`);
console.log(` new_types: ${summarize(j.data_model?.new_types, 400)}`);
console.log(` risks: ${summarize(j.risks, 300)}`);
console.log(` why>summary: ${summarize(j.why_this_beats_summarization, 300)}`);
}
// Write full JSON to disk so we can inspect later.
const outPath = `/home/profit/lakehouse/data/_kb/consensus_reducer_design_${Date.now().toString(36)}.json`;
await Bun.write(outPath, JSON.stringify(results, null, 2));
console.log(`\nfull responses → ${outPath}`);
}
await main();
tests/real-world/enrich_prd_pipeline.ts
-528
View File
@@ -1,528 +0,0 @@
// Real-world architecture stress test — 6 iterations of the full pipeline
// against the PRD as a corpus. Goal: prove at scale what Phase 21
// promised (context continuation + tree-split), plus Phase 19
// compounding across iterations.
//
// Run: bun run tests/real-world/enrich_prd_pipeline.ts
//
// No mocks. No skipped layers. On any error, the test triggers
// cloud-rescue rather than fail — it's the architecture's job to
// recover. The test FAILS only if we can't complete 6 iterations.
import { readFile, writeFile, mkdir } from "node:fs/promises";
import { createHash } from "node:crypto";
const PRD_PATH = "/home/profit/lakehouse/docs/PRD.md";
const CHUNK_SIZE = 800; // chars per chunk — ~200 tokens
const CHUNK_OVERLAP = 120;
const TOP_K_RETRIEVE = 12; // chunks per iteration — pulled up to force overflow
const CONTEXT_BUDGET_CHARS = 4000; // tight budget — forces tree-split on every iter
const INJECT_FAIL_ON_ITER = 3; // force the TASK-retry loop on iter 3
// Continuation controls (per-cloud-call) — used for output-overflow.
// Separate from the task-retry loop (per-task) — that handles errors
// across attempts.
const PRIMARY_MAX_TOKENS = 150; // tight — forces truncation
const CONTINUATION_MAX_TOKENS = 300; // each continuation doubles headroom
const MAX_CONTINUATIONS = 6; // max stitch pieces per cloud call
// Task-level retry loop (J's clarification, 2026-04-22):
// When a TASK errors, retry the whole task up to 6 times. Each
// retry gets prior attempts' failures injected as learning context,
// so attempt N+1 is informed by what N failed at. The loop caps at
// 6 to avoid infinite spinning on genuinely unsolvable tasks.
const MAX_TASK_RETRIES = 6;
// To FORCE the retry loop on iter INJECT_FAIL_ON_ITER, cycle through
// 5 deliberately-invalid models + 1 valid one. Attempts 1-5 will
// 502/404 from Ollama Cloud; attempt 6 finally succeeds. Proves the
// loop fires all 6 with compounding failure context.
const FORCE_RETRY_MODEL_SEQUENCE = [
"deliberately-invalid-model-attempt-1",
"deliberately-invalid-model-attempt-2",
"deliberately-invalid-model-attempt-3",
"deliberately-invalid-model-attempt-4",
"deliberately-invalid-model-attempt-5",
"gpt-oss:20b", // 6th attempt succeeds
];
const GATEWAY = "http://localhost:3100";
const SIDECAR = "http://localhost:3200";
const CLOUD_MODEL = "gpt-oss:120b";
const RESCUE_MODEL = "gpt-oss:20b"; // fallback local cloud model via sidecar
const RUN_NONCE = Date.now().toString(36);
const OUT_DIR = `/home/profit/lakehouse/tests/real-world/runs/${RUN_NONCE}`;
// The 6 progressively-compounding questions. #6 explicitly requires
// synthesis across prior 5 answers.
const QUESTIONS: string[] = [
"Summarize the Lakehouse project's one-paragraph thesis: what problem does it solve, what's the unique approach?",
"How does Phase 19 playbook memory turn successful fills into a signal that boosts future rankings?",
"Explain the role of Phase 24 observer in the learning loop — what does it see, and what does it feed into?",
"What's the VRAM-aware profile swap mechanism in Phase 17, and why does it matter for multi-model serving?",
"How do Phase 25 validity windows and Phase 27 playbook versioning interact when a schema drifts?",
"Synthesize the prior 5 answers: how do the pieces (playbook memory, observer, profile swap, validity windows, versioning) compose into a system that measurably gets smarter over time? Cite specific prior answers.",
];
type Chunk = { id: string; text: string; embedding: number[]; offset: number };
interface IterationResult {
iteration: number;
question: string;
retrieval_top_k: number;
context_chars_before_budget: number;
tree_split_fired: boolean;
cloud_calls_total: number;
continuation_retries: number;
rescue_triggered: boolean;
// Task-level retry telemetry
task_attempts_made: number; // how many attempts fired (1 = first succeeded)
task_retry_history: Array<{ n: number; model: string; error: string }>;
playbook_id: string | null;
tokens_prompt: number;
tokens_completion: number;
citations_from_prior_iterations: string[];
duration_ms: number;
answer_preview: string;
errors_recovered: string[];
}
function log(msg: string) { console.log(`[enrich] ${msg}`); }
function sleep(ms: number) { return new Promise(r => setTimeout(r, ms)); }
function cosine(a: number[], b: number[]): number {
let dot = 0, na = 0, nb = 0;
for (let i = 0; i < a.length; i++) { dot += a[i] * b[i]; na += a[i] * a[i]; nb += b[i] * b[i]; }
if (na === 0 || nb === 0) return 0;
return dot / (Math.sqrt(na) * Math.sqrt(nb));
}
function hash(s: string): string { return createHash("sha256").update(s).digest("hex").slice(0, 10); }
async function embedBatch(texts: string[]): Promise<number[][]> {
// Sidecar /embed accepts a list. On partial failure, retry individually.
const r = await fetch(`${SIDECAR}/embed`, {
method: "POST", headers: { "content-type": "application/json" },
body: JSON.stringify({ texts }),
signal: AbortSignal.timeout(120000),
});
if (!r.ok) throw new Error(`embed batch ${r.status}: ${await r.text()}`);
const j: any = await r.json();
return j.embeddings;
}
function chunkText(text: string): Array<{ text: string; offset: number }> {
const out: Array<{ text: string; offset: number }> = [];
let i = 0;
while (i < text.length) {
const end = Math.min(i + CHUNK_SIZE, text.length);
const slice = text.slice(i, end).trim();
if (slice.length > 50) out.push({ text: slice, offset: i });
if (end >= text.length) break;
i = end - CHUNK_OVERLAP;
}
return out;
}
async function chat(opts: {
provider: "ollama" | "ollama_cloud",
model: string,
messages: Array<{ role: string; content: string }>,
max_tokens: number,
think: boolean,
}): Promise<{ content: string; prompt_tokens: number; completion_tokens: number; finish_reason: string }> {
const r = await fetch(`${GATEWAY}/v1/chat`, {
method: "POST", headers: { "content-type": "application/json" },
body: JSON.stringify({ ...opts }),
signal: AbortSignal.timeout(180000),
});
if (!r.ok) throw new Error(`/v1/chat ${r.status}: ${await r.text()}`);
const j: any = await r.json();
return {
content: j.choices?.[0]?.message?.content ?? "",
prompt_tokens: j.usage?.prompt_tokens ?? 0,
completion_tokens: j.usage?.completion_tokens ?? 0,
finish_reason: j.choices?.[0]?.finish_reason ?? "?",
};
}
// ─── Tree-split over oversized chunk set ──────────────────────────
async function treeSplitSummarize(
chunks: Chunk[],
question: string,
): Promise<{ scratchpad: string; cloud_calls: number }> {
// Shard into groups fitting within half the budget each.
const perShard = Math.max(1, Math.floor((CONTEXT_BUDGET_CHARS / 2) / CHUNK_SIZE));
const shards: Chunk[][] = [];
for (let i = 0; i < chunks.length; i += perShard) {
shards.push(chunks.slice(i, i + perShard));
}
log(` tree-split: ${chunks.length} chunks → ${shards.length} shards of up to ${perShard}`);
let scratchpad = "";
let cloud_calls = 0;
for (let si = 0; si < shards.length; si++) {
const shard = shards[si];
const shardText = shard.map(c => `[chunk @${c.offset}]\n${c.text}`).join("\n\n");
const userMsg = `Question: ${question}\n\nShard ${si + 1}/${shards.length} of source material:\n\n${shardText}\n\nScratchpad so far:\n${scratchpad || "(empty)"}\n\nUpdate the scratchpad: extract only facts from THIS shard that help answer the question. Be terse. No prose.`;
const r = await chat({
provider: "ollama_cloud",
model: CLOUD_MODEL,
messages: [
{ role: "system", content: "You maintain a concise factual scratchpad across multiple shards of source text. No prose outside the scratchpad. Each shard, append ≤80 words of relevant facts." },
{ role: "user", content: userMsg },
],
max_tokens: 500,
think: false,
});
cloud_calls += 1;
scratchpad += `\n--- shard ${si + 1} notes ---\n${r.content.trim()}`;
if (scratchpad.length > CONTEXT_BUDGET_CHARS) {
// truncate oldest halves
scratchpad = scratchpad.slice(-CONTEXT_BUDGET_CHARS);
log(` tree-split: scratchpad truncated to ${scratchpad.length} chars`);
}
}
return { scratchpad, cloud_calls };
}
// ─── Continuable generate — up to max_continuations stitches ──────
//
// Two failure modes handled:
// A) Empty response — typically thinking model burned the budget
// on hidden reasoning. Retry with 2× max_tokens.
// B) Truncated response (finish_reason=length) — answer got cut off
// mid-sentence. Pass the partial back as scratchpad and ask the
// model to continue from where it stopped.
//
// Stitching: keep appending content across retries; prompt_tokens and
// completion_tokens accumulate; finish_reason reflects the LAST call.
// Loop exits on the first call that finishes cleanly (stop) with
// non-empty content, OR when retries hit the cap.
async function generateContinuable(
opts: Parameters<typeof chat>[0] & { max_continuations?: number },
): Promise<{ content: string; prompt_tokens: number; completion_tokens: number; continuation_retries: number; finish_reason: string }> {
const maxCont = opts.max_continuations ?? 1;
let total = await chat(opts);
let retries = 0;
while (retries < maxCont && (total.content.length === 0 || total.finish_reason === "length")) {
retries += 1;
const mode = total.content.length === 0 ? "empty" : "truncated";
log(` continuation retry ${retries}/${maxCont} (${mode}: finish=${total.finish_reason}, content=${total.content.length} chars)`);
// Continuation prompt — branch on failure mode:
// empty → retry with 2× tokens, same prompt (thinking budget)
// length → pass the partial as assistant turn, ask to continue
const continuationMessages = total.content.length === 0
? opts.messages
: [
...opts.messages,
{ role: "assistant", content: total.content },
{ role: "user", content: "Continue from exactly where you stopped. Do not repeat. Finish the answer." },
];
const continued = await chat({
...opts,
max_tokens: CONTINUATION_MAX_TOKENS,
messages: continuationMessages,
});
total = {
content: total.content + continued.content,
prompt_tokens: total.prompt_tokens + continued.prompt_tokens,
completion_tokens: total.completion_tokens + continued.completion_tokens,
finish_reason: continued.finish_reason,
};
}
return { ...total, continuation_retries: retries };
}
// ─── Single iteration: retrieve → budget-check → chat → seed ─────
async function runIteration(
iteration: number,
question: string,
allChunks: Chunk[],
priorPlaybookIds: string[],
priorAnswers: string[],
): Promise<IterationResult> {
const started = Date.now();
const errorsRecovered: string[] = [];
log(`iter ${iteration}: "${question.slice(0, 70)}..."`);
// 1. Embed the question
const qEmb = (await embedBatch([question]))[0];
// 2. Retrieve top-K chunks by cosine
const scored = allChunks
.map(c => ({ c, score: cosine(qEmb, c.embedding) }))
.sort((a, b) => b.score - a.score)
.slice(0, TOP_K_RETRIEVE);
const chunks = scored.map(x => x.c);
log(` retrieved top ${chunks.length} chunks (score range ${scored[0].score.toFixed(3)} .. ${scored[scored.length - 1].score.toFixed(3)})`);
// 3. Context budget check — tree-split if over
const contextChars = chunks.map(c => c.text).join("\n\n").length;
let contextForPrompt: string;
let treeSplit = false;
let cloudCallsTotal = 0;
if (contextChars > CONTEXT_BUDGET_CHARS) {
treeSplit = true;
log(` context ${contextChars} chars > budget ${CONTEXT_BUDGET_CHARS} → tree-split`);
const { scratchpad, cloud_calls } = await treeSplitSummarize(chunks, question);
contextForPrompt = `Distilled scratchpad from ${chunks.length} source chunks (too large to fit directly):\n${scratchpad}`;
cloudCallsTotal += cloud_calls;
} else {
contextForPrompt = chunks.map(c => `[chunk @${c.offset}]\n${c.text}`).join("\n\n");
}
// 4. Seed prompt with prior iteration answers (real compounding).
// Not just IDs — the model needs the CONTENT to synthesize.
let citationBlock = "";
let citationsReceived: string[] = [];
if (priorPlaybookIds.length > 0 && priorAnswers.length > 0) {
const lines = priorAnswers.map((ans, i) => {
const pid = priorPlaybookIds[i]?.slice(0, 12) ?? "unknown";
// Trim each prior answer to ~400 chars so we don't blow budget
return `[pb:${pid}] iter ${i + 1} answer:\n${ans.slice(0, 400)}\n`;
});
citationBlock = `\n\n═══ PRIOR ITERATIONS (compounding context) ═══\n${lines.join("\n")}═══ end prior iterations ═══\n\nYour answer MUST cite specific prior iterations using [pb:ID] notation when drawing on them. Synthesis questions require explicit cross-iteration reasoning.`;
citationsReceived = priorPlaybookIds.slice();
}
// 5. TASK-LEVEL RETRY LOOP — per J's clarification 2026-04-22.
// Try the task up to MAX_TASK_RETRIES times. Each retry:
// a) Picks a model (normally CLOUD_MODEL; on INJECT_FAIL_ON_ITER,
// cycles through 5 invalid models + 1 valid to force full loop)
// b) Injects prior attempt errors as learning context
// c) If the attempt succeeds (non-empty, >100 chars), loop exits
// d) Otherwise, records failure and tries again with the learning
//
// Cap at 6 so we don't spin forever on unsolvable tasks.
let result: { content: string; prompt_tokens: number; completion_tokens: number; continuation_retries: number; finish_reason: string } | null = null;
let rescueTriggered = false;
const taskAttemptHistory: Array<{ n: number; model: string; error: string }> = [];
const forceRetries = iteration === INJECT_FAIL_ON_ITER;
if (forceRetries) log(` FORCING TASK-RETRY LOOP — iter ${iteration} will cycle through 5 invalid models + 1 valid`);
for (let attempt = 1; attempt <= MAX_TASK_RETRIES; attempt++) {
const modelForAttempt = forceRetries
? FORCE_RETRY_MODEL_SEQUENCE[attempt - 1]
: CLOUD_MODEL;
// Compose a prior-attempts learning block for attempts 2+
const learningBlock = taskAttemptHistory.length > 0
? `\n\n═══ PRIOR ATTEMPTS THIS TASK (do NOT repeat these failures; adjust approach) ═══\n${taskAttemptHistory.map(a => `Attempt ${a.n} (model ${a.model}) failed: ${a.error.slice(0, 160)}`).join("\n")}\n═══ end prior attempts ═══\n`
: "";
log(` task attempt ${attempt}/${MAX_TASK_RETRIES}: model=${modelForAttempt}${learningBlock ? " [with prior-failure context]" : ""}`);
try {
const r = await generateContinuable({
provider: "ollama_cloud",
model: modelForAttempt,
messages: [
{ role: "system", content: "You answer questions about the Lakehouse PRD using only the provided source material and prior iteration answers. Be specific. Cite chunk offsets OR [pb:ID] markers. Write a detailed 250-word answer." },
{ role: "user", content: `Question: ${question}\n\nSource material:\n${contextForPrompt}${citationBlock}${learningBlock}` },
],
max_tokens: PRIMARY_MAX_TOKENS,
think: false,
max_continuations: MAX_CONTINUATIONS,
});
cloudCallsTotal += 1 + r.continuation_retries;
if (r.content && r.content.length > 100) {
// Acceptable answer — exit loop
result = r;
if (attempt > 1) {
log(` task attempt ${attempt} SUCCEEDED (${r.content.length} chars) after ${attempt - 1} prior failures`);
rescueTriggered = true;
}
break;
}
// Thin response — count as failure with learning signal
const err = `thin-answer: ${r.content.length} chars, finish=${r.finish_reason}`;
taskAttemptHistory.push({ n: attempt, model: modelForAttempt, error: err });
errorsRecovered.push(`attempt ${attempt}: ${err}`);
} catch (e) {
const err = (e as Error).message;
taskAttemptHistory.push({ n: attempt, model: modelForAttempt, error: err });
errorsRecovered.push(`attempt ${attempt}: ${err.slice(0, 120)}`);
cloudCallsTotal += 1;
}
}
// Last-ditch: if all 6 task attempts failed, try the local fallback
// once more so we at least return SOMETHING. This is the "don't get
// caught in a loop, accept best-so-far" rule J stated explicitly.
if (!result) {
errorsRecovered.push(`all ${MAX_TASK_RETRIES} task attempts failed — local fallback`);
rescueTriggered = true;
try {
result = await generateContinuable({
provider: "ollama",
model: "qwen3.5:latest",
messages: [{ role: "user", content: `Q: ${question}\n\n${contextForPrompt.slice(0, 4000)}` }],
max_tokens: 300,
think: false,
max_continuations: 2,
});
cloudCallsTotal += 1 + result.continuation_retries;
} catch (e) {
// Absolute last resort — fabricate a skeleton result
result = {
content: `[task failed after ${MAX_TASK_RETRIES} retries + local fallback: ${(e as Error).message}]`,
prompt_tokens: 0,
completion_tokens: 0,
continuation_retries: 0,
finish_reason: "error",
};
}
}
if (result.content.length === 0) {
errorsRecovered.push("even rescue returned empty — last-ditch local fallback");
rescueTriggered = true;
result = await generateContinuable({
provider: "ollama",
model: "qwen3.5:latest",
messages: [{ role: "user", content: `Q: ${question}\n\n${contextForPrompt.slice(0, 4000)}` }],
max_tokens: 300,
think: false,
});
cloudCallsTotal += 1;
}
// 6. Seed playbook with the answer
let playbook_id: string | null = null;
try {
const ts = new Date().toISOString();
const seedOp = `TEST: enrich_prd_run_${RUN_NONCE} iter${iteration} in Corpus, PRD`;
const r = await fetch(`${GATEWAY}/vectors/playbook_memory/seed`, {
method: "POST", headers: { "content-type": "application/json" },
body: JSON.stringify({
operation: seedOp,
approach: `q="${question.slice(0, 80)}" context_chars=${contextChars} tree_split=${treeSplit}`,
context: result.content.slice(0, 600),
endorsed_names: [`iter${iteration}_${RUN_NONCE}`],
append: true,
}),
signal: AbortSignal.timeout(15000),
});
if (r.ok) {
const j: any = await r.json();
playbook_id = j.outcome?.playbook_id ?? null;
} else {
errorsRecovered.push(`seed ${r.status}: ${(await r.text()).slice(0, 100)}`);
}
} catch (e) {
errorsRecovered.push(`seed exception: ${(e as Error).message}`);
}
return {
iteration,
question,
retrieval_top_k: chunks.length,
context_chars_before_budget: contextChars,
tree_split_fired: treeSplit,
cloud_calls_total: cloudCallsTotal,
continuation_retries: result.continuation_retries,
rescue_triggered: rescueTriggered,
task_attempts_made: taskAttemptHistory.length + 1, // +1 for the successful attempt
task_retry_history: taskAttemptHistory,
playbook_id,
tokens_prompt: result.prompt_tokens,
tokens_completion: result.completion_tokens,
citations_from_prior_iterations: citationsReceived,
duration_ms: Date.now() - started,
answer_preview: result.content.slice(0, 500),
errors_recovered: errorsRecovered,
};
}
async function main() {
await mkdir(OUT_DIR, { recursive: true });
log(`run nonce: ${RUN_NONCE}`);
log(`output dir: ${OUT_DIR}`);
// ─── Phase 1: load, chunk, embed the PRD ───────────────────────
log(`loading PRD from ${PRD_PATH}`);
const prd = await readFile(PRD_PATH, "utf8");
log(`PRD: ${prd.length} chars, ${prd.split("\n").length} lines`);
const raw_chunks = chunkText(prd);
log(`chunked into ${raw_chunks.length} pieces (size ${CHUNK_SIZE}, overlap ${CHUNK_OVERLAP})`);
// Embed in batches of 32 to avoid sidecar overload
const allChunks: Chunk[] = [];
const BATCH = 32;
const t0 = Date.now();
for (let i = 0; i < raw_chunks.length; i += BATCH) {
const batch = raw_chunks.slice(i, i + BATCH);
const embs = await embedBatch(batch.map(b => b.text));
for (let j = 0; j < batch.length; j++) {
allChunks.push({
id: hash(batch[j].text),
text: batch[j].text,
embedding: embs[j].map(x => Number(x)),
offset: batch[j].offset,
});
}
log(` embedded ${allChunks.length}/${raw_chunks.length}`);
}
log(`embedded all ${allChunks.length} chunks in ${((Date.now() - t0) / 1000).toFixed(1)}s`);
// ─── Phase 2: 6 iterations ─────────────────────────────────────
const results: IterationResult[] = [];
const priorIds: string[] = [];
const priorAnswers: string[] = [];
for (let i = 1; i <= QUESTIONS.length; i++) {
const q = QUESTIONS[i - 1];
const r = await runIteration(i, q, allChunks, priorIds, priorAnswers);
results.push(r);
if (r.playbook_id) priorIds.push(r.playbook_id);
priorAnswers.push(r.answer_preview);
log(` → iter ${i}: ${r.errors_recovered.length} errors recovered, ${r.continuation_retries} continuations, tree-split=${r.tree_split_fired}, rescue=${r.rescue_triggered}, ${r.duration_ms}ms`);
await writeFile(`${OUT_DIR}/iter_${i}.json`, JSON.stringify(r, null, 2));
}
// Check whether iter 6 actually cited prior pb:IDs in its answer.
// Playbook IDs look like `pb-seed-<hex>` so the regex needs to allow
// hyphens + letters inside the brackets, not just hex chars.
const iter6 = results[5];
const citationsHonored = iter6 ? (iter6.answer_preview.match(/\[pb:[\w-]+\]/gi)?.length ?? 0) : 0;
// ─── Phase 3: summary ──────────────────────────────────────────
const summary = {
run_nonce: RUN_NONCE,
ran_at: new Date().toISOString(),
prd_chars: prd.length,
prd_chunks: allChunks.length,
iterations: results.length,
total_cloud_calls: results.reduce((s, r) => s + r.cloud_calls_total, 0),
total_continuation_retries: results.reduce((s, r) => s + r.continuation_retries, 0),
total_errors_recovered: results.reduce((s, r) => s + r.errors_recovered.length, 0),
tree_splits_fired: results.filter(r => r.tree_split_fired).length,
rescues_triggered: results.filter(r => r.rescue_triggered).length,
iter6_received_prior_ids: results[5]?.citations_from_prior_iterations.length ?? 0,
iter6_actually_cited_in_answer: citationsHonored,
iter3_task_attempts: results[2]?.task_attempts_made ?? 0,
iter3_task_retries: results[2]?.task_retry_history.length ?? 0,
max_task_attempts_any_iter: Math.max(...results.map(r => r.task_attempts_made)),
total_duration_ms: results.reduce((s, r) => s + r.duration_ms, 0),
overall: results.length === 6 && results.every(r => r.playbook_id !== null) ? "PASS" : "PARTIAL",
};
await writeFile(`${OUT_DIR}/summary.json`, JSON.stringify(summary, null, 2));
log("");
log(`══════ SUMMARY ${summary.overall} ══════`);
log(` 6 iterations, ${summary.total_cloud_calls} cloud calls, ${summary.total_errors_recovered} errors recovered`);
log(` tree-splits: ${summary.tree_splits_fired}/6 continuations: ${summary.total_continuation_retries} rescues: ${summary.rescues_triggered}`);
log(` iter 6 received ${summary.iter6_received_prior_ids} prior IDs, cited ${summary.iter6_actually_cited_in_answer} [pb:...] markers in its answer`);
log(` iter 3 task-retry loop: ${summary.iter3_task_attempts} attempts (${summary.iter3_task_retries} prior-failure retries before success)`);
log(` total duration: ${(summary.total_duration_ms / 1000).toFixed(1)}s`);
log("");
for (const r of results) {
const flags = [
r.tree_split_fired ? "tree-split" : "",
r.continuation_retries > 0 ? `cont=${r.continuation_retries}` : "",
r.rescue_triggered ? "rescued" : "",
r.errors_recovered.length > 0 ? `err=${r.errors_recovered.length}` : "",
].filter(Boolean).join(" ");
log(` iter ${r.iteration}: ${r.tokens_prompt}+${r.tokens_completion} tok, ${r.duration_ms}ms ${flags ? `[${flags}]` : ""}`);
}
log("");
log(`artifacts: ${OUT_DIR}/{iter_1..6.json, summary.json}`);
process.exit(summary.overall === "PASS" ? 0 : 1);
}
main().catch(e => { console.error("[enrich] fatal:", e); process.exit(2); });
tests/real-world/hard_task_escalation.ts
-267
View File
@@ -1,267 +0,0 @@
// Hard-task escalation test. The task is deliberately constructed so
// that a local 7B model (qwen3.5:latest) will miss at least one of the
// validation rules. Watch the escalation ladder:
// 1. qwen3.5:latest (local 7B) — likely fails
// 2. qwen3:latest (local 7B) — likely fails differently
// 3. gpt-oss:20b (cloud 20B) — may fail
// 4. gpt-oss:120b (cloud 120B) — should succeed
// 5. gpt-oss:120b w/ prior-attempt errors injected — retry with context
// 6. absolute last ditch: return best-so-far with failure annotation
//
// Each attempt:
// - Calls the model via /v1/chat
// - Validates the output against a strict rubric
// - On fail: records the specific rubric violations + the partial
// output, injects both into the next attempt's prompt as "here's
// what's wrong, fix it specifically"
// - On success: exit loop
//
// Run: bun run tests/real-world/hard_task_escalation.ts
import { writeFile, mkdir } from "node:fs/promises";
const GATEWAY = "http://localhost:3100";
const OUT_DIR = `/home/profit/lakehouse/tests/real-world/runs/hard_task_${Date.now().toString(36)}`;
const MAX_ATTEMPTS = 6;
// The hard task. Specific enough that a small model will miss at
// least one rule. Not purely knowledge-based — it's a code-generation
// task with strict structural constraints.
const TASK = `Write a complete Rust async function with the EXACT signature:
pub async fn check_drift_batched(refs: Vec<DocRef>) -> Result<Vec<String>, String>
It must:
1. Group refs by tool name (case-insensitive — use .to_ascii_lowercase())
2. Issue parallel HTTP GET requests to http://localhost:3900/docs/{tool}/diff?since={snippet_hash}
3. Use reqwest and a JoinSet/Semaphore to cap concurrent in-flight requests at 4
4. On HTTP 5xx, retry with exponential backoff: sleep 250ms, then 500ms, then 1000ms, then give up on that tool
5. Parse the response JSON: {"drifted": bool, ...}. Return a Vec<String> of tool names where drifted == true
6. All errors bubble via ? or Result — NO .unwrap(), NO .expect(), NO panic!()
7. Include rustdoc /// comments on the function and each internal helper
Assume this struct is already imported:
pub struct DocRef { pub tool: String, pub snippet_hash: Option<String>, pub version_seen: String }
Output ONLY the Rust code. No prose, no markdown fences, no explanation. Start directly with the /// doc comment.`;
// Escalation ladder — small-local → large-local → cloud → specialist
// cloud → trillion-param cloud. Corrected 2026-04-22 per J:
// gpt-oss:20b is LOCAL (ollama list confirms 13 GB on disk), and the
// final escalation tier should be kimi-k2:1t (the biggest model
// we have access to on Ollama Cloud).
const LADDER: Array<{ provider: "ollama" | "ollama_cloud"; model: string; note: string }> = [
{ provider: "ollama", model: "qwen3.5:latest", note: "local 7B" },
{ provider: "ollama", model: "qwen3:latest", note: "local 7B (different) " },
{ provider: "ollama", model: "gpt-oss:20b", note: "local 20B" }, // FIXED: local, not cloud
{ provider: "ollama_cloud", model: "gpt-oss:120b", note: "cloud 120B" },
{ provider: "ollama_cloud", model: "devstral-2:123b", note: "cloud 123B (coding specialist)" },
// NOTE 2026-04-22 — J wanted Kimi as the last escalation but Kimi
// K2.5/K2.6 both return "this model requires a subscription" on our
// current Ollama Cloud key. mistral-large-3:675b is the biggest
// model actually provisioned on this key (verified via direct curl
// to ollama.com/api/generate). Upgrade path: Ollama Cloud Pro →
// swap this line to kimi-k2.5 or kimi-k2.6:cloud.
{ provider: "ollama_cloud", model: "mistral-large-3:675b", note: "cloud 675B (biggest available on current key; kimi-k2.x needs pro subscription)" },
];
// Validation rubric — the answer must pass all of these to be accepted.
interface RubricResult {
passed: boolean;
violations: string[];
passed_rules: string[];
}
function validate(code: string): RubricResult {
const violations: string[] = [];
const passed: string[] = [];
const check = (rule: string, ok: boolean) => { ok ? passed.push(rule) : violations.push(rule); };
check("has pub async fn check_drift_batched signature",
/pub\s+async\s+fn\s+check_drift_batched\s*\(/.test(code));
check("takes Vec<DocRef> argument",
/refs\s*:\s*Vec\s*<\s*DocRef\s*>/.test(code));
check("returns Result<Vec<String>, String>",
/Result\s*<\s*Vec\s*<\s*String\s*>\s*,\s*String\s*>/.test(code));
check("uses reqwest",
/\breqwest\b/i.test(code));
check("references JoinSet or Semaphore for concurrency",
/\bJoinSet\b|\bSemaphore\b/i.test(code));
check("bounds concurrency at 4",
/\b4\b/.test(code) && (/Semaphore\s*::\s*new\s*\(\s*4\b/.test(code) || /permits\s*:\s*4\b/.test(code) || /limit\s*:\s*4\b/.test(code) || /max\s*:\s*4\b/.test(code) || /capacity\s*:\s*4\b/.test(code)));
// Exponential backoff — models express this several ways. Accept
// any recognizable doubling pattern starting at 250ms. 2026-04-22:
// devstral-2:123b wrote `retry_delay *= 2` which my earlier regex
// rejected even though the code is correct. Broadening rubric to
// match all idiomatic doubling forms.
const hasSeed250 = /Duration\s*::\s*from_millis\s*\(\s*250\b/.test(code)
|| /millis\s*\(\s*250\b/.test(code);
const hasDoublingPattern = /250\s*\*\s*2/.test(code) // 250 * 2^n literal
|| /<<\s*\d+/.test(code) // bit-shift
|| /\.pow\s*\(/.test(code) // 2u32.pow(attempt)
|| /\*=\s*2\b/.test(code) // delay *= 2 ← was missing
|| /\*\s*2\s*;/.test(code) // delay = delay * 2;
|| /saturating_mul\s*\(\s*2\b/.test(code); // saturating doubling
check("has 250ms backoff seed",
hasSeed250);
check("reaches 500ms backoff (literal or doubling from 250)",
/Duration\s*::\s*from_millis\s*\(\s*500\b/.test(code)
|| /millis\s*\(\s*500\b/.test(code)
|| (hasSeed250 && hasDoublingPattern));
check("reaches 1000ms backoff (literal or doubling to 1000)",
/Duration\s*::\s*from_millis\s*\(\s*1000\b/.test(code)
|| /millis\s*\(\s*1000\b/.test(code)
|| (hasSeed250 && hasDoublingPattern));
check("case-insensitive tool grouping (to_ascii_lowercase)",
/to_ascii_lowercase|to_lowercase/.test(code));
check("NO .unwrap() — all errors bubble via ?",
!/\.unwrap\s*\(\s*\)/.test(code));
check("NO .expect(...) — all errors bubble via ?",
!/\.expect\s*\(/.test(code));
check("NO panic!() / unimplemented!() / todo!()",
!/\bpanic!\s*\(|\bunimplemented!\s*\(|\btodo!\s*\(/.test(code));
check("has rustdoc /// comments",
/\/\/\//.test(code));
check("reasonable length (> 500 chars)",
code.length > 500);
return { passed: violations.length === 0, violations, passed_rules: passed };
}
function log(msg: string) { console.log(`[hard] ${msg}`); }
async function chat(opts: {
provider: "ollama" | "ollama_cloud",
model: string,
prompt: string,
}): Promise<{ content: string; error?: string }> {
try {
const r = await fetch(`${GATEWAY}/v1/chat`, {
method: "POST", headers: { "content-type": "application/json" },
body: JSON.stringify({
provider: opts.provider,
model: opts.model,
messages: [{ role: "user", content: opts.prompt }],
max_tokens: 2500,
temperature: 0.2,
think: false,
}),
signal: AbortSignal.timeout(240000),
});
if (!r.ok) return { content: "", error: `/v1/chat ${r.status}: ${(await r.text()).slice(0, 300)}` };
const j: any = await r.json();
return { content: j.choices?.[0]?.message?.content ?? "" };
} catch (e) {
return { content: "", error: (e as Error).message };
}
}
interface AttemptRecord {
n: number;
provider: string;
model: string;
duration_ms: number;
content_chars: number;
error: string | null;
rubric_violations: string[];
rubric_passed: string[];
accepted: boolean;
}
function extractCode(raw: string): string {
// Strip common fence wrappers
const m = raw.match(/```(?:rust)?\s*\n([\s\S]*?)```/);
if (m) return m[1].trim();
return raw.trim();
}
async function main() {
await mkdir(OUT_DIR, { recursive: true });
log(`output: ${OUT_DIR}`);
log(`task: ${TASK.slice(0, 120)}...`);
log("");
const attempts: AttemptRecord[] = [];
let acceptedCode: string | null = null;
for (let i = 0; i < MAX_ATTEMPTS; i++) {
const n = i + 1;
const rung = LADDER[i] ?? LADDER[LADDER.length - 1];
// Build the prompt: base task + prior failures' learning blocks
let priorLearning = "";
if (attempts.length > 0) {
priorLearning = `\n\n═══ PRIOR ATTEMPTS FAILED. Fix these exact issues: ═══\n`;
for (const a of attempts) {
priorLearning += `Attempt ${a.n} (${a.provider}/${a.model}, ${a.content_chars} chars) violations:\n`;
for (const v of a.rubric_violations) priorLearning += ` - ${v}\n`;
if (a.error) priorLearning += ` [error: ${a.error.slice(0, 120)}]\n`;
}
priorLearning += `═══ end prior attempts ═══\n\nDO NOT repeat the above violations. Address each one explicitly.`;
}
log(`attempt ${n}/${MAX_ATTEMPTS}: ${rung.provider}::${rung.model}${priorLearning ? " [w/ learning]" : ""}`);
const t0 = Date.now();
const r = await chat({ provider: rung.provider, model: rung.model, prompt: TASK + priorLearning });
const dur = Date.now() - t0;
const code = extractCode(r.content);
const rubric = code ? validate(code) : { passed: false, violations: ["empty response"], passed_rules: [] };
const record: AttemptRecord = {
n,
provider: rung.provider,
model: rung.model,
duration_ms: dur,
content_chars: code.length,
error: r.error ?? null,
rubric_violations: rubric.violations,
rubric_passed: rubric.passed_rules,
accepted: rubric.passed,
};
attempts.push(record);
log(` → ${dur}ms, ${code.length} chars, ${rubric.passed_rules.length} rules passed / ${rubric.violations.length} failed${r.error ? `, err: ${r.error.slice(0, 80)}` : ""}`);
for (const v of rubric.violations.slice(0, 5)) log(` ✗ ${v}`);
await writeFile(`${OUT_DIR}/attempt_${n}.txt`, code);
await writeFile(`${OUT_DIR}/attempt_${n}.json`, JSON.stringify(record, null, 2));
if (rubric.passed) {
log(` ✅ ACCEPTED on attempt ${n}`);
acceptedCode = code;
break;
}
}
const summary = {
task: TASK.slice(0, 200),
total_attempts: attempts.length,
accepted: acceptedCode !== null,
accepted_on_attempt: acceptedCode ? attempts.findIndex(a => a.accepted) + 1 : null,
escalation_path: attempts.map(a => `${a.provider}/${a.model}`),
per_attempt_pass_counts: attempts.map(a => a.rubric_passed.length),
per_attempt_violation_counts: attempts.map(a => a.rubric_violations.length),
total_duration_ms: attempts.reduce((s, a) => s + a.duration_ms, 0),
};
await writeFile(`${OUT_DIR}/summary.json`, JSON.stringify(summary, null, 2));
log("");
log(`═══ RESULT ═══`);
log(`attempts: ${summary.total_attempts}`);
log(`accepted: ${summary.accepted} ${summary.accepted ? `on attempt ${summary.accepted_on_attempt}` : ""}`);
log(`escalation path:`);
for (const [i, a] of attempts.entries()) {
const mark = a.accepted ? "✅" : "❌";
log(` ${mark} attempt ${i + 1}: ${a.provider}/${a.model} — ${a.rubric_passed.length}/${a.rubric_passed.length + a.rubric_violations.length} rules passed, ${a.duration_ms}ms`);
}
log("");
log(`total time: ${(summary.total_duration_ms / 1000).toFixed(1)}s`);
log(`artifacts: ${OUT_DIR}/{attempt_1..N.{txt,json}, summary.json}`);
process.exit(summary.accepted ? 0 : 1);
}
main().catch(e => { console.error("[hard] fatal:", e); process.exit(2); });
tests/real-world/nine_consecutive_audits.ts
-186
View File
@@ -1,186 +0,0 @@
// Nine-consecutive audit runner — empirical test of the predictive-
// compounding property. Runs the audit pipeline 9 times against the
// same PR (each time with a new diff from Gitea), captures the
// verdict + audit_lessons state after each run, and reports whether
// the KB stabilizes or drifts.
//
// What we expect (favorable compounding):
// - signature_count grows sublinearly (same patterns recur, so
// distinct-signature count stabilizes fast)
// - verdict settles on a stable value after run 2-3 (first audit
// establishes baseline, rest repeat)
// - confidence stays LOW for all signatures (same PR repeatedly)
// - NO new recurring findings fire because confidence < 0.3 on
// same-PR noise (kb_index rating policy)
//
// What would indicate drift (the thing we want to prove DOESN'T happen):
// - signature_count grows linearly — each run produces new signatures
// - verdict oscillates (block → approve → block ...)
// - confidence inflates — kb_index rating escalates on repeated runs
//
// Run: bun run tests/real-world/nine_consecutive_audits.ts
import { readFile, writeFile } from "node:fs/promises";
import { join } from "node:path";
import { aggregate } from "../../auditor/kb_index.ts";
import { getPrSnapshot } from "../../auditor/gitea.ts";
import { auditPr } from "../../auditor/audit.ts";
const REPO = "/home/profit/lakehouse";
const AUDIT_LESSONS = `${REPO}/data/_kb/audit_lessons.jsonl`;
const VERDICTS_DIR = `${REPO}/data/_auditor/verdicts`;
const POLL_INTERVAL_MS = 5_000;
const RUNS = Number(process.env.LH_AUDIT_RUNS ?? 9);
const TARGET_PR = Number(process.env.LH_AUDIT_PR ?? 8);
const SKIP_INFERENCE = process.env.LH_AUDITOR_SKIP_INFERENCE !== "0";
const RESET_KB = process.env.LH_RESET_KB === "1";
async function waitForVerdict(prNum: number, sha: string, deadlineMs: number): Promise<any> {
const short = sha.slice(0, 12);
const path = join(VERDICTS_DIR, `${prNum}-${short}.json`);
const start = Date.now();
while (Date.now() - start < deadlineMs) {
try {
const raw = await readFile(path, "utf8");
return JSON.parse(raw);
} catch { /* not yet */ }
await Bun.sleep(POLL_INTERVAL_MS);
}
throw new Error(`no verdict file after ${deadlineMs}ms: ${path}`);
}
async function captureAggState(): Promise<{ sig_count: number; max_count: number; max_confidence: number; top3: Array<{ sig: string; count: number; conf: number; summary: string }> }> {
const agg = await aggregate<any>(AUDIT_LESSONS, {
keyFn: (r) => r?.signature,
scopeFn: (r) => (r?.pr_number !== undefined ? `pr-${r.pr_number}` : undefined),
});
const list = Array.from(agg.values()).sort((a, b) => b.count - a.count);
const recurring = list.filter(r => r.count >= 2);
const recurringMaxCount = recurring.length > 0 ? Math.max(...recurring.map(a => a.count)) : 0;
const recurringMaxConf = recurring.length > 0 ? Math.max(...recurring.map(a => a.confidence)) : 0;
return {
sig_count: list.length,
max_count: list[0]?.count ?? 0,
max_confidence: recurringMaxConf,
recurring_max_count: recurringMaxCount,
top3: list.slice(0, 3).map(a => ({
sig: a.signature,
count: a.count,
conf: a.confidence,
summary: a.representative_summary.slice(0, 80),
})),
};
}
interface RunRecord {
run: number;
sha: string;
verdict_overall: string;
findings_total: number;
findings_block: number;
findings_warn: number;
findings_info: number;
audit_duration_ms: number;
claims_total: number;
claims_empirical: number;
kb_sig_count_after: number;
kb_max_count_after: number;
kb_max_confidence_after: number;
kb_recurring_max_count: number;
}
async function main() {
console.log(`[nine] target PR: #${TARGET_PR}`);
console.log(`[nine] runs: ${RUNS}`);
console.log(`[nine] skip_inference: ${SKIP_INFERENCE}`);
console.log(`[nine] reset_kb: ${RESET_KB}`);
console.log(`[nine] audit_lessons.jsonl: ${AUDIT_LESSONS}`);
if (RESET_KB) {
console.log("[nine] clearing audit_lessons.jsonl for clean test...");
await writeFile(AUDIT_LESSONS, "");
}
console.log("");
const pr = await getPrSnapshot(TARGET_PR);
console.log(`[nine] PR #${pr.number}: "${pr.title}" (head=${pr.head_sha.slice(0, 12)})`);
console.log(`[nine] files in diff: ${pr.files.length}`);
console.log("");
const baseline = await captureAggState();
console.log(`[nine] baseline: sig_count=${baseline.sig_count} max_count=${baseline.max_count} max_conf=${baseline.max_confidence.toFixed(2)}`);
console.log("");
const records: RunRecord[] = [];
for (let n = 1; n <= RUNS; n++) {
const t0 = Date.now();
console.log(`─── run ${n}/${RUNS} ───`);
const verdict = await auditPr(pr, {
dry_run: true,
skip_dynamic: true,
skip_inference: SKIP_INFERENCE,
});
console.log(` sha ${verdict.head_sha.slice(0, 12)}`);
const after = await captureAggState();
const rec: RunRecord = {
run: n,
sha: verdict.head_sha.slice(0, 12),
verdict_overall: String(verdict.overall),
findings_total: Number(verdict.metrics?.findings_total ?? 0),
findings_block: Number(verdict.metrics?.findings_block ?? 0),
findings_warn: Number(verdict.metrics?.findings_warn ?? 0),
findings_info: Number(verdict.metrics?.findings_info ?? 0),
audit_duration_ms: Number(verdict.metrics?.audit_duration_ms ?? 0),
claims_total: Number(verdict.metrics?.claims_total ?? 0),
claims_empirical: Number(verdict.metrics?.claims_empirical ?? 0),
kb_sig_count_after: after.sig_count,
kb_max_count_after: after.max_count,
kb_max_confidence_after: after.max_confidence,
kb_recurring_max_count: after.recurring_max_count,
};
records.push(rec);
console.log(` verdict=${rec.verdict_overall} findings=${rec.findings_total} (b=${rec.findings_block} w=${rec.findings_warn})`);
console.log(` kb after: sig=${rec.kb_sig_count_after} max_count=${rec.kb_max_count_after} recurring_max=${rec.kb_recurring_max_count} max_conf=${rec.kb_max_confidence_after.toFixed(2)}`);
console.log(` elapsed: ${((Date.now() - t0) / 1000).toFixed(1)}s`);
console.log("");
}
console.log("═══ FINAL ═══");
console.log("run | verdict | find | block warn info | dur_s | kb_sig max_count max_conf");
for (const r of records) {
console.log(
` ${String(r.run).padStart(1)} | ${r.verdict_overall.padEnd(16)} | ${String(r.findings_total).padStart(4)} | ${String(r.findings_block).padStart(5)} ${String(r.findings_warn).padStart(5)} ${String(r.findings_info).padStart(5)} | ${(r.audit_duration_ms / 1000).toFixed(1).padStart(5)} | ${String(r.kb_sig_count_after).padStart(6)} ${String(r.kb_max_count_after).padStart(9)} ${r.kb_max_confidence_after.toFixed(2)}`,
);
}
console.log("");
console.log("═══ COMPOUNDING PROPERTY ═══");
const sigDelta = records[records.length - 1].kb_sig_count_after - baseline.sig_count;
const maxConf = records[records.length - 1].kb_max_confidence_after;
const recurringMax = records[records.length - 1].kb_recurring_max_count;
console.log(` signatures added over ${RUNS} runs: ${sigDelta}`);
console.log(` max recurring count after run ${RUNS}: ${recurringMax} (same-PR recurrences per signature)`);
console.log(` max confidence after run ${RUNS}: ${maxConf.toFixed(2)} (expect LOW — same-PR should not inflate)`);
const verdictSet = new Set(records.map(r => r.verdict_overall));
if (verdictSet.size === 1) {
console.log(` verdict stable: all ${RUNS} runs returned '${[...verdictSet][0]}' ✓`);
} else {
console.log(` verdict oscillated across runs: ${[...verdictSet].join(" | ")} ✗`);
}
if (maxConf < 0.6 && recurringMax < 5) {
console.log(` confidence policy holding: same-PR noise stays below escalation threshold ✓`);
} else {
console.log(` ⚠ cross-cutting pattern detected (conf=${maxConf.toFixed(2)}, recurring=${recurringMax}) — kb_index policy escalated`);
}
const jsonOut = `${REPO}/tests/real-world/runs/nine_consecutive_${Date.now().toString(36)}.json`;
await Bun.write(jsonOut, JSON.stringify({ target_pr: TARGET_PR, baseline, records }, null, 2));
console.log("");
console.log(` report: ${jsonOut}`);
}
main().catch(e => { console.error("[nine] fatal:", e); process.exit(1); });