Build the first working bounded research agent

Build only what the claim requires: agent stops at max_steps; every claim cites a KB id or abstains. Prefer boring, deterministic code over frameworks you cannot yet explain. Run the path twice; identical output on this fixture is a feature, not a lack of creativity.

export async function run(input: unknown, deps: Dependencies) {
  const started = performance.now();
  const validated = deps.parseInput(input);
  const raw = await deps.execute(validated); // alternate one model decision with at most one validated tool execution per loop iteration
  const value = deps.parseOutput(raw);
  return { value, durationMs: Math.round(performance.now() - started) };
}

Expected evidence: The agent searches once, cites kb://returns, and stops before MAX_AGENT_STEPS.. Read each printed intermediate as part of the argument that the path works—not as decoration.

Narrate the journey from raw input to result for a single example from fixture KB with 4 notes + max_steps=3. If you cannot name an intermediate, the implementation is still too opaque for this lab. Only after this path is solid should you generalize data sources or UI.

While implementing the bounded research agent, print or log at least three intermediates that map to the claim (agent stops at max_steps; every claim cites a KB id or abstains). Good intermediates are values a teammate could recompute with a calculator or diff. Bad intermediates are framework traces you cannot explain.

Re-run with fixture KB with 4 notes + max_steps=3 twice. If the second run differs, either the path is nondeterministic (document the seed) or you have hidden global state—both are lab bugs until named.

Stop adding features once the path supports search a local knowledge base then produce a cited answer within a step budget. Extra UI, extra tools, or extra models belong in later chapters. The mastery bar for this page is simply: a deterministic end-to-end path with intermediates.

At the implementation stage for agents-in-code, the job is narrower than finishing a product demo. You are creating one progressive evidence piece about fixture KB with 4 notes + max_steps=3 that later pages inherit without redefining success. Keep that fixture small enough to inspect by hand, keep outputs copy-pasteable as text, and refuse to narrate this baseline as if it were a production SLA: single-shot LLM answer without search.

For this page specifically, success looks like a deterministic path with printed intermediates while still centering the user decision to search a local knowledge base then produce a cited answer within a step budget. If you cannot point to a file, command, or assertion that proves that for the bounded research agent, stay on this page instead of advancing.

Glossary: tool · Glossary: structured output · Cheatsheet: production ops signals

Previous · Next