@companion-ai/feynman 0.5.36 → 0.5.38

package/.feynman/SYSTEM.md

@@ -6,7 +6,7 @@ Evidence:
 - Evidence over fluency. Prefer papers, official documentation, datasets, code, and direct experimental results over commentary.
 - Separate observations from inferences, and state uncertainty explicitly.
 - When a claim depends on recent literature or unstable facts, use tools before answering.
-- For papers, use the alpha tools (`alpha_search`, `alpha_get_paper`, `alpha_ask_paper`, `alpha_read_code`) and `feynman_science_database_search`. In a shell, call `feynman alpha ...`, not a bare global `alpha`.
+- For papers, use the alpha tools (`alpha_search`, `alpha_get_paper`, `alpha_ask_paper`, `alpha_read_code`) and `feynman_science_database_search`. In a shell, call `feynman alpha ...`, not a bare global `alpha`. On macOS and Linux run Python as `python3`; `python` is often missing there.
 - For current topics (products, releases, pricing, benchmarks, docs, regulations, anything latest/current/recent), use `web_search` and `fetch_content` first. Never answer a current question from paper search alone. For mixed topics, use both.
 - Cite title, year, and a direct URL or identifier; prefer arXiv or alphaXiv links with the arXiv ID. Source-based answers end with a Sources section of direct URLs.
 

package/RELEASES.md

@@ -6,6 +6,16 @@ GitHub release notes are generated from the matching `## vX.Y.Z` section in this
 
 ## Unreleased
 
+## v0.5.38 - 2026-10-12
+
+- **`/deepresearch` no longer waits for a "yes" after its plan.** About 1 in 5 people who started it never answered the approval question, so the research never ran. It now writes the plan, summarizes it, and continues, as `/lit` does. To check the plan first, say so in the request, for example `/deepresearch <topic>, show me the plan first`.
+- **A failed model call says what went wrong.** When the provider rejects a request, Feynman now names the cause and the fix under the error: a model your account can't use (pick another with `/model`), no credits, a rate limit or used-up quota, rejected credentials (`feynman model login <provider>`), or a temporary outage. Most failed `/lit` and `/audit` runs ended on one of these. Feynman bugs go to github.com/Companion-Inc/feynman/issues; Pi's `/bug` reports to Pi's developers.
+- Research runs on macOS and Linux no longer fail on `python: command not found`; Feynman tells the model to run `python3`.
+
+## v0.5.37 - 2026-10-12
+
+- **A sign-in that came back without a code says what to do.** When the browser returned to `feynman model login` or `feynman setup` without an authorization code (seen mostly on Windows), Feynman printed only "Missing authorization code". It now gives the same advice as for `invalid_grant`: sign out of the provider's site or use a private window, then retry, and for ChatGPT it suggests `feynman model login openai-codex`.
+
 ## v0.5.36 - 2026-10-12
 
 - **A new user gets to a working model instead of a dead end.** Launching `feynman` with no model opens a menu to set one up.

package/dist/model/commands.js

@@ -742,11 +742,12 @@ export function getOrCreateDeviceId(authPath) {
     writeFileSync(settingsPath, `${JSON.stringify({ ...settings, deviceId }, null, 2)}\n`, "utf8");
     return deviceId;
 }
-// A provider answers invalid_grant when it rejects the authorization code,
-// usually because a stale session in the browser interfered with the sign-in.
-// Signing out there (or using a private window) and retrying fixes it.
+// A provider answers invalid_grant when it rejects the authorization code, and
+// the browser comes back with no code when the sign-in was denied or lost a
+// redirect; a stale browser session usually causes both. Signing out there (or
+// using a private window) and retrying fixes it.
 export function describeOAuthLoginFailure(providerId, error) {
-    if (!(error instanceof Error) || !error.message.includes("invalid_grant"))
+    if (!(error instanceof Error) || !/invalid_grant|Missing authorization code/.test(error.message))
         return error;
     const alternative = providerId === "openai" ? " Or sign in with `feynman model login openai-codex` instead." : "";
     return new Error(`${error.message}\nThe sign-in was rejected, usually because of a stale browser session. Sign out of the provider's site in your browser, or use a private window, then run \`feynman model login ${providerId}\` again.${alternative}`);

package/extensions/research-tools/model-error-hint.ts

@@ -0,0 +1,36 @@
+import type { ExtensionAPI } from "@earendil-works/pi-coding-agent";
+
+// Most failed research runs end on a provider error (quota, a key, a model the
+// plan lacks). Pi follows them with its /bug hint, which reports to Pi's
+// developers; this names the cause and the fix instead.
+export function modelErrorHint(provider: string, error: string): string | undefined {
+	const login = `feynman model login ${provider}`;
+	if (/model_not_supported|requested model is not supported|not available to new users|does not support tools/i.test(error)) {
+		return `Your ${provider} account can't use this model. Pick another with /model (or \`feynman model set\`).`;
+	}
+	if (/\b402\b|credit|billing|balance is insufficient|payment required/i.test(error)) {
+		return `Your ${provider} account is out of credits. Add credits, or switch provider with /model.`;
+	}
+	if (/\b429\b|quota|rate.?limit|usage limit|resource.?exhausted|capacity|too many requests/i.test(error)) {
+		return `${provider} is rate-limiting this account or its quota is used up. Wait and retry, or switch model with /model.`;
+	}
+	if (/\b40[13]\b|invalid.{0,20}(api.?key|x-api-key|token)|authentication|unauthori[sz]ed|invalid_grant|refresh token/i.test(error)) {
+		return `${provider} rejected your credentials. Run \`${login}\` again, or check the API key you set.`;
+	}
+	if (/\b50[0-4]\b|overloaded|high demand|unavailable|timed out|timeout|terminated|connection error|fetch failed|ECONNRESET/i.test(error)) {
+		return `${provider} failed or timed out; this is usually temporary. Retry, or switch model with /model.`;
+	}
+	return undefined;
+}
+
+export function registerModelErrorHint(pi: ExtensionAPI): void {
+	pi.on("message_end", (event, ctx) => {
+		const message = event.message;
+		if (message.role !== "assistant" || message.stopReason !== "error" || !ctx.hasUI) return;
+		const hint = modelErrorHint(message.provider, message.errorMessage ?? "");
+		ctx.ui.notify(
+			`${hint ?? `${message.provider} returned an error.`} If this looks like a Feynman bug, report it at https://github.com/Companion-Inc/feynman/issues (Pi's /bug goes to Pi's developers).`,
+			"warning",
+		);
+	});
+}

package/extensions/research-tools.ts

@@ -8,6 +8,7 @@ import { installFeynmanHeader } from "./research-tools/header.js";
 import { registerHelpCommand } from "./research-tools/help.js";
 import { registerHuggingFaceTools } from "./research-tools/huggingface.js";
 import { registerLaunchNotice } from "./research-tools/launch-notice.js";
+import { registerModelErrorHint } from "./research-tools/model-error-hint.js";
 import { registerInitCommand, registerOutputsCommand } from "./research-tools/project.js";
 import { registerServiceTierControls } from "./research-tools/service-tier.js";
 import { registerScienceDatabaseTools } from "./research-tools/science-databases.js";
@@ -29,6 +30,7 @@ export default function researchTools(pi: ExtensionAPI): void {
 	registerHelpCommand(pi);
 	registerInitCommand(pi);
 	registerLaunchNotice(pi);
+	registerModelErrorHint(pi);
 	registerOutputsCommand(pi);
 	registerServiceTierControls(pi);
 	registerScienceDatabaseTools(pi);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@companion-ai/feynman",
-  "version": "0.5.36",
+  "version": "0.5.38",
   "description": "The open source AI research agent.",
   "license": "MIT",
   "type": "module",

package/prompts/deepresearch.md

@@ -17,7 +17,7 @@ Every run must leave these files on disk:
 - `outputs/<slug>.md` or `papers/<slug>.md`
 - `outputs/<slug>.provenance.md` or `papers/<slug>.provenance.md`
 
-Before the plan is approved, only the plan file may exist; do not create placeholder drafts, final outputs, or provenance files. After the user approves the plan, if any capability fails, continue in degraded mode and still write a blocked or partial final output and provenance sidecar. Never end with chat-only output after plan approval. Never end with only an explanation in chat after plan approval. Use `Verification: BLOCKED` when verification could not be completed.
+Do not create placeholder drafts, final outputs, or provenance files before their stage. If any capability fails, continue in degraded mode and still write a blocked or partial final output and provenance sidecar. Never end with chat-only output or with only an explanation in chat. Use `Verification: BLOCKED` when verification could not be completed.
 
 ## Step 1: Plan
 
@@ -31,11 +31,7 @@ Create `outputs/.plans/<slug>.md` immediately. The plan must include:
 
 Make the scale decision before assigning owners in the plan. If the topic is a narrow "what is X" explainer, the plan must use lead-owned direct search tasks only; do not allocate researcher subagents in the task ledger.
 
-After writing the plan, stop and ask for explicit confirmation before gathering evidence. Summarize the plan briefly and ask:
-
-`Proceed with this deep research plan? Reply "yes" to continue, or tell me what to change.`
-
-Do not run searches, fetch sources, spawn subagents, draft, cite, review, or deliver final artifacts until the user confirms. If the user requests changes, update `outputs/.plans/<slug>.md` first, then ask for confirmation again.
+After writing the plan, summarize it in a few lines and continue immediately in the same turn with the next tool call; never end a turn with a message that only announces what you will do next. Stop for confirmation only when the user explicitly asked to review the plan first; then ask `Proceed with this deep research plan? Reply "yes" to continue, or tell me what to change.`, and if they request changes, update `outputs/.plans/<slug>.md` before asking again.
 
 ## Step 2: Scale