@companion-ai/feynman 0.5.14 → 0.5.16

package/.feynman/agents/researcher.md

@@ -3,7 +3,6 @@ name: researcher
 description: Gather primary evidence across papers, web sources, repos, docs, and local artifacts.
 thinking: high
 tools: read, write, edit, bash, grep, find, ls, web_search, fetch_content, get_search_content, feynman_science_database_search, hf_dataset_info, hf_repo_files, hf_repo_read_file
-output: research.md
 defaultProgress: true
 async: true
 ---
@@ -94,7 +93,7 @@ Numbered list matching the evidence table:
 - If you were assigned multiple questions, track them explicitly in the file and mark each as `done`, `blocked`, or `needs follow-up`. Do not silently skip questions.
 
 ## Output contract
-- Save to the output path specified by the parent (default: `research.md`).
+- Save to the path the parent specifies, relative to the workspace. If none is given, return the complete artifact in your final response.
 - Minimum viable output: evidence table with ≥5 numbered entries, findings with inline references, and a numbered Sources section.
 - Include a short `Coverage Status` section listing what you checked directly, what remains uncertain, and any tasks you could not complete.
 - Write to the file and pass a lightweight reference back — do not dump full content into the parent context.

package/.feynman/agents/reviewer.md

@@ -2,7 +2,6 @@
 name: reviewer
 description: Run tough but constructive internal research critique of an AI research artifact.
 thinking: high
-output: review.md
 defaultProgress: true
 ---
 
@@ -88,5 +87,5 @@ Reference the weakness/question IDs from Part 1 so annotations link back to the
 - End with a `Sources` section containing direct URLs for anything additionally inspected during review.
 
 ## Output contract
-- Save the main artifact to the output path specified by the parent (default: `review.md`).
+- Save to the path the parent specifies, relative to the workspace. If none is given, return the complete artifact in your final response.
 - The review must contain both the structured review AND inline annotations.

package/.feynman/agents/verifier.md

@@ -3,7 +3,6 @@ name: verifier
 description: Post-process a draft to add inline citations and verify every source URL.
 thinking: medium
 tools: read, bash, grep, find, ls, write, edit, web_search, fetch_content, get_search_content, feynman_science_database_search
-output: cited.md
 defaultProgress: true
 async: true
 ---
@@ -57,6 +56,6 @@ Before saving the final document, scan for:
 For each item, verify that it maps to a source URL, research note, raw artifact path, or script path. If not, remove it or replace it with a TODO. Add a short `Removed Unsupported Claims` section only when you remove material.
 
 ## Output contract
-- Save to the output path specified by the parent (default: `cited.md`).
+- Save to the path the parent specifies, relative to the workspace. If none is given, return the complete artifact in your final response.
 - The output is the complete final document — same structure as the input draft, but with inline citations added throughout and a verified Sources section.
 - Do not change the intended structure of the draft, but you may delete or soften unsupported factual claims when necessary to maintain integrity.

package/.feynman/agents/writer.md

@@ -3,7 +3,6 @@ name: writer
 description: Turn research notes into clear, structured briefs and drafts.
 thinking: medium
 tools: read, bash, grep, find, ls, write, edit
-output: draft.md
 defaultProgress: true
 ---
 
@@ -52,5 +51,5 @@ Unresolved issues, disagreements between sources, gaps in evidence.
 - Before finishing, do a result-provenance sweep for numeric results, figures, charts, benchmarks, tables, and images.
 
 ## Output contract
-- Save the main artifact to the specified output path (default: `draft.md`).
+- Save to the path the parent specifies, relative to the workspace. If none is given, return the complete artifact in your final response.
 - Focus on clarity, structure, and evidence traceability.

package/RELEASES.md

@@ -6,6 +6,22 @@ GitHub release notes are generated from the matching `## vX.Y.Z` section in this
 
 ## Unreleased
 
+## v0.5.16 - 2026-09-30
+
+- **Subagent research files land in the workspace.** When a workflow gave a subagent a relative output path such as `outputs/.drafts/<slug>-research-web.md`, pi-subagents saved it under `~/.feynman/sessions/subagent-artifacts/outputs/<run>/` instead, so the lead agent could not find the research notes, citations, or verification it asked for (about one in seven installs running research workflows hit this in the last week). Relative subagent outputs now resolve against the workspace, the researcher, reviewer, verifier, and writer no longer default to generic file names such as `research.md`, and `/deepresearch` and `/lit` name every subagent file under `outputs/.drafts/`. If you set `singleRunOutputBaseDir` in `~/.feynman/agent/extensions/subagent/config.json`, Feynman keeps your value.
+- `feynman packages remove <preset>` (also `rm` and `uninstall`) removes an optional package such as `memory`.
+- On Windows, Feynman no longer fails to start when another Feynman start is copying `keybindings.json` (`EBUSY: resource busy or locked`).
+- pi-web-access 0.34.0: `fetch_content` falls back to your other fetch providers on Cloudflare "Just a moment..." pages, web tools start available on models such as DeepSeek without costing a prompt-cache miss, and its undici moves to 8.11.2.
+- Security: three brace-expansion advisories published on 2026-09-29 affect the 5.0.9 that Pi 0.99.1's shrinkwrap locks, which Feynman cannot override. Pi uses it only for package filters and model patterns from your own settings and flags; the next Pi release that ships 5.0.12 replaces it. pi-subagents has still not released its undici fix.
+
+## v0.5.15 - 2026-09-29
+
+- **Pi 0.99.1, pinned.** Feynman accepted any Pi version, so npm installs picked up Pi 0.99.1 the day it was released, untested, and Pi 0.99 warns on every launch about how Feynman declared it. Feynman now runs on Pi 0.99.1, which adds MCP server support (`/mcp`), declared the way Pi asks, and pinned so new Pi releases arrive only after they pass Feynman's checks. If npm has `legacy-peer-deps` on, it skips Pi; `feynman` now says to reinstall with `--legacy-peer-deps=false` instead of failing on a missing package.
+- **npm installs use a patched undici.** pi-subagents pins undici 8.10.0, which ten advisories published on 2026-09-29 affect (plus GHSA-3wwx-pv8p-q78v), and npm hoisted that copy for every package in an npm install. Feynman now depends on a patched undici directly, so only pi-subagents keeps 8.10.0 until it releases its bump (nicobailon/pi-subagents#2548); the standalone installers were already patched. Correction to the 0.5.14 notes: pi-subagents' background runner installs its undici as the process-wide fetch and WebSocket, so the advisories do apply to background subagent runs.
+- **`feynman model set` explains a model it cannot use.** "Model not available in Pi auth storage" gave no reason. It now says which applies: the provider has no credentials (with the `feynman model login` command to run), a local provider in `models.json` has no `apiKey` (Pi hides providers without one; a placeholder such as `"local"` is enough), or the ID is unknown (with close matches, for example `gpt-5.6` suggests `openai/gpt-5.6-terra`).
+- **Parallel OpenAlex semantic searches no longer fail.** OpenAlex allows one semantic search per second, and parallel searches got `429 Too Many Requests`: in a live test of six at once, five failed. They now run one after another at that pace, and a 429 that asks for a short wait is retried once. Six parallel searches now all succeed.
+- `feynman model set` with no model opens a picker of the available models instead of failing. `feynman model help` lists the model commands, an unknown model subcommand points to it, and `-h` works like `--help`.
+
 ## v0.5.14 - 2026-09-28
 
 - **The Windows installer no longer fails on slow networks (#309).** When usage telemetry could not reach PostHog quickly, the PostHog library printed a timeout to stderr, and Windows PowerShell treats stderr from the installer's `feynman --help` check as a failure. Each telemetry send now has a 1.5 s budget and ends quietly, so no command waits on telemetry (on a stalled network `feynman --help` took 11 s and printed two errors; it now takes about 2 s and prints nothing). The installers also run their checks with telemetry off and judge them by exit code.

package/bin/feynman.js

@@ -40,4 +40,15 @@ if (process.argv.length === 3 && process.argv[2] === "--version") {
   process.exit(0);
 }
 
+// Pi is a peer dependency, as Pi requires of extension packages, and npm skips
+// peers when legacy-peer-deps is on.
+try {
+  import.meta.resolve("@earendil-works/pi-coding-agent");
+  import.meta.resolve("@earendil-works/pi-ai");
+} catch {
+  console.error("feynman could not find its Pi runtime (@earendil-works/pi-coding-agent), which npm skips when legacy-peer-deps is on.");
+  console.error("Reinstall with: npm install -g @companion-ai/feynman --legacy-peer-deps=false");
+  process.exit(1);
+}
+
 await import(pathToFileURL(resolve(here, "..", "dist", "index.js")).href);

package/dist/bootstrap/home.js

@@ -1,5 +1,5 @@
 import { createHash } from "node:crypto";
-import { copyFileSync, existsSync, readFileSync, renameSync, rmSync } from "node:fs";
+import { constants, copyFileSync, existsSync, readFileSync, renameSync, rmSync } from "node:fs";
 import { resolve } from "node:path";
 import { getBootstrapStatePath } from "../config/paths.js";
 const LEGACY_SYNCED_SCOPES = { themes: "themes", agents: "agents", skills: "skills" };
@@ -41,6 +41,12 @@ export function ensureFeynmanAgentDir(appRoot, home, agentDir) {
     }
     const keybindingsPath = resolve(agentDir, "keybindings.json");
     if (!existsSync(keybindingsPath)) {
-        copyFileSync(resolve(appRoot, ".feynman", "config", "keybindings.json"), keybindingsPath);
+        try {
+            copyFileSync(resolve(appRoot, ".feynman", "config", "keybindings.json"), keybindingsPath, constants.COPYFILE_EXCL);
+        }
+        catch {
+            // Another Feynman start may be writing it (Windows reports EBUSY). Pi
+            // uses its default keys without the file, and the next start copies it.
+        }
     }
 }

package/dist/cli.js

@@ -18,11 +18,11 @@ import { verifyAlphaAuthStatus } from "./alpha-auth-status.js";
 import { ensureFeynmanAgentDir } from "./bootstrap/home.js";
 import { ensureFeynmanHome, getDefaultSessionDir, getFeynmanAgentDir, getFeynmanHome } from "./config/paths.js";
 import { getLastPiStderr, launchPiChat, runPi } from "./pi/launch.js";
-import { installPiPackage, listOptionalPackagePresets, normalizeOptionalPackagePresetName, resolvePackageSource, updatePiPackages, } from "./pi/packages.js";
+import { installPiPackage, removePiPackage, listOptionalPackagePresets, normalizeOptionalPackagePresetName, resolvePackageSource, updatePiPackages, } from "./pi/packages.js";
 import { canonicalizeModelSpec, ensureFeynmanSettings, normalizeThinkingLevel, readJson, } from "./pi/settings.js";
 import { BUNDLED_PI_PACKAGES, buildPiEnv } from "./pi/runtime.js";
 import { getConfiguredServiceTier, normalizeServiceTier, setConfiguredServiceTier } from "./model/service-tier.js";
-import { authenticateModelProvider, getCurrentModelSpec, isLocalModelProvider, loginModelProvider, logoutModelProvider, printModelList, setDefaultModelSpec, } from "./model/commands.js";
+import { authenticateModelProvider, getCurrentModelSpec, isLocalModelProvider, loginModelProvider, logoutModelProvider, printModelList, selectDefaultModel, setDefaultModelSpec, } from "./model/commands.js";
 import { buildModelStatusSnapshotFromRecords, getAuthenticatedModelRecords, isProClassModelSpec, getSupportedModelRecords, } from "./model/catalog.js";
 import { clearSearchConfig, printSearchStatus, setSearchProvider } from "./search/commands.js";
 import { fetchLatestFeynmanVersion, getFeynmanUpgradeLines, isNewerVersion } from "./system/self-update.js";
@@ -189,10 +189,15 @@ async function handleModelCommand(subcommand, args, feynmanSettingsPath, feynman
     }
     if (subcommand === "set") {
         const spec = args[0];
-        if (!spec) {
+        if (spec) {
+            await setDefaultModelSpec(feynmanSettingsPath, feynmanAuthPath, spec);
+        }
+        else if (process.stdin.isTTY && process.stdout.isTTY) {
+            await selectDefaultModel(feynmanSettingsPath, feynmanAuthPath);
+        }
+        else {
             throw new Error("Usage: feynman model set <provider/model|provider:model>");
         }
-        await setDefaultModelSpec(feynmanSettingsPath, feynmanAuthPath, spec);
         return;
     }
     if (subcommand === "tier") {
@@ -214,7 +219,14 @@ async function handleModelCommand(subcommand, args, feynmanSettingsPath, feynman
         console.log(`Service tier set to ${tier}`);
         return;
     }
-    throw new Error(`Unknown model command: ${subcommand}`);
+    if (subcommand === "help") {
+        printSection("Model Management");
+        for (const command of cliCommandSections.find((section) => section.title === "Model Management")?.commands ?? []) {
+            printHelpLine(command.usage, command.description);
+        }
+        return;
+    }
+    throw new Error(`Unknown model command: ${subcommand}\nRun \`feynman model help\` to see model commands.`);
 }
 async function handleUpdateCommand(piOptions, feynmanVersion, source) {
     const latestFeynmanVersionPromise = fetchLatestFeynmanVersion();
@@ -255,18 +267,27 @@ async function handlePackagesCommand(subcommand, args, piOptions) {
         printInfo(`Install with: feynman packages install <${optionalPresets.map((preset) => preset.name).join("|")}>`);
         return;
     }
-    if (subcommand !== "install") {
-        throw new Error(`Unknown packages command: ${subcommand}`);
+    const remove = subcommand === "remove" || subcommand === "rm" || subcommand === "uninstall";
+    if (subcommand !== "install" && !remove) {
+        throw new Error(`Unknown packages command: ${subcommand}\nUse: feynman packages list, install <preset>, or remove <preset>.`);
     }
     const target = args[0];
     if (!target) {
-        throw new Error(`Usage: feynman packages install <${listOptionalPackagePresets().map((preset) => preset.name).join("|")}>`);
+        throw new Error(`Usage: feynman packages ${remove ? "remove" : "install"} <${listOptionalPackagePresets().map((preset) => preset.name).join("|")}>`);
     }
     const presetName = normalizeOptionalPackagePresetName(target);
     if (!presetName) {
         throw new Error(`Unknown package preset: ${target}`);
     }
     const source = resolvePackageSource(presetName);
+    if (remove) {
+        if (!configuredSources.has(source)) {
+            console.log(`${source} is not installed`);
+            return;
+        }
+        process.exitCode = await removePiPackage(piOptions, source);
+        return;
+    }
     if (configuredSources.has(source)) {
         console.log(`${source} already installed`);
         return;
@@ -436,7 +457,7 @@ async function runMain(input) {
             doctor: { type: "boolean" },
             export: { type: "string" },
             fork: { type: "string" },
-            help: { type: "boolean" },
+            help: { type: "boolean", short: "h" },
             version: { type: "boolean" },
             "alpha-login": { type: "boolean" },
             "alpha-logout": { type: "boolean" },
@@ -614,7 +635,7 @@ async function runMain(input) {
         const modelRuntime = await createModelRuntime(feynmanAuthPath);
         const canonicalModelSpec = canonicalizeModelSpec(requestedExplicitModelSpec, modelRuntime);
         if (!canonicalModelSpec) {
-            throw new Error(`Unknown model: ${requestedExplicitModelSpec}`);
+            throw new Error(`Unknown model: ${requestedExplicitModelSpec}. Run \`feynman model list\` to see available models.`);
         }
         explicitModelSpec = canonicalModelSpec;
     }

package/dist/model/commands.js

@@ -5,7 +5,7 @@ import { readJson } from "../pi/settings.js";
 import { promptChoice, promptSelect, promptText } from "../setup/prompts.js";
 import { openUrl } from "../system/open-url.js";
 import { printInfo, printSection, printSuccess, printWarning } from "../ui/terminal.js";
-import { buildModelStatusSnapshotFromRecords, chooseRecommendedModel, getAuthenticatedModelRecords, getAvailableModelRecords, getSupportedModelRecords, } from "./catalog.js";
+import { buildModelStatusSnapshotFromRecords, chooseRecommendedModel, getAuthenticatedModelRecords, getAvailableModelRecords, getSupportedModelRecords, isProClassModelSpec, } from "./catalog.js";
 import { MODEL_API_KEY_PROVIDERS } from "./api-key-providers.js";
 import { createModelRegistry, createModelRuntime, getModelsJsonPath } from "./registry.js";
 import { upsertProviderBaseUrl, upsertProviderConfig } from "./models-json.js";
@@ -821,10 +821,32 @@ export async function logoutModelProvider(authPath, providerId) {
     await modelRuntime.logout(provider.id);
     printSuccess(`Model provider logout complete: ${provider.id}`);
 }
+// Explains why `feynman model set <spec>` found no usable model.
+async function describeUnavailableModelSpec(authPath, spec) {
+    const input = spec.trim().replace(/^([^/:]+):(.+)$/, "$1/$2").toLowerCase();
+    const known = (await getSupportedModelRecords(authPath)).find((model) => `${model.provider}/${model.id}`.toLowerCase() === input || model.id.toLowerCase() === input);
+    if (known) {
+        const knownSpec = `${known.provider}/${known.id}`;
+        if (isProClassModelSpec(knownSpec)) {
+            return `Pro-class model disabled: ${knownSpec}. Choose an approved research model.`;
+        }
+        if (isLocalModelProvider(authPath, known.provider)) {
+            return `${knownSpec} has no API key. Pi only lists a local provider when it has one: add "apiKey": "local" to the ${known.provider} provider in ${getModelsJsonPath(authPath)}.`;
+        }
+        return `${knownSpec} has no credentials. Run \`feynman model login ${known.provider}\` first.`;
+    }
+    const modelId = input.slice(input.lastIndexOf("/") + 1);
+    const matches = (await getAvailableModelRecords(authPath))
+        .map((model) => `${model.provider}/${model.id}`)
+        .filter((candidate) => candidate.toLowerCase().includes(modelId))
+        .slice(0, 5);
+    const suggestion = matches.length > 0 ? ` Did you mean ${matches.join(", ")}?` : "";
+    return `Unknown model: ${spec}.${suggestion} Run \`feynman model list\` to see available models.`;
+}
 export async function setDefaultModelSpec(settingsPath, authPath, spec) {
     const resolvedSpec = await resolveAvailableModelSpec(authPath, spec);
     if (!resolvedSpec) {
-        throw new Error(`Model not available in Pi auth storage: ${spec}. Run \`feynman model list\` first.`);
+        throw new Error(await describeUnavailableModelSpec(authPath, spec));
     }
     const [provider, ...rest] = resolvedSpec.split("/");
     const modelId = rest.join("/");
@@ -834,6 +856,22 @@ export async function setDefaultModelSpec(settingsPath, authPath, spec) {
     writeFileSync(settingsPath, JSON.stringify(settings, null, 2) + "\n", "utf8");
     printSuccess(`Research default model set to ${resolvedSpec}`);
 }
+export async function selectDefaultModel(settingsPath, authPath) {
+    const status = await collectModelStatus(settingsPath, authPath);
+    if (status.availableModels.length === 0) {
+        printWarning("No authenticated Pi models are currently available.");
+        for (const line of status.guidance) {
+            printInfo(line);
+        }
+        return;
+    }
+    const spec = await promptSelect("Choose the default research model:", status.availableModels.map((value) => ({
+        value,
+        label: value,
+        hint: value === status.current ? "current" : value === status.recommended ? "recommended" : undefined,
+    })), status.current ?? status.recommended);
+    await setDefaultModelSpec(settingsPath, authPath, spec);
+}
 export async function runModelSetup(settingsPath, authPath) {
     let status = await collectModelStatus(settingsPath, authPath);
     while (status.availableModels.length === 0) {

package/dist/pi/packages.js

@@ -27,6 +27,9 @@ export function resolvePackageSource(name) {
 export async function installPiPackage(options, source) {
     return runPi(options, ["install", source], buildPiEnv(options));
 }
+export async function removePiPackage(options, source) {
+    return runPi(options, ["remove", source], buildPiEnv(options));
+}
 export async function updatePiPackages(options, source) {
     return runPi(options, source ? ["update", source] : ["update", "--extensions"], buildPiEnv(options));
 }

package/dist/pi/settings.js

@@ -95,7 +95,13 @@ function prepareSubagentDefaults(settingsPath) {
             throw new Error(`Invalid subagent config at ${path}: ${key} must be a boolean. The file was not changed.`);
         }
     }
-    if (missions.enabled !== undefined && config.fleetView !== undefined && config.asyncByDefault !== undefined) {
+    if (config.singleRunOutputBaseDir !== undefined && typeof config.singleRunOutputBaseDir !== "string") {
+        throw new Error(`Invalid subagent config at ${path}: singleRunOutputBaseDir must be a string. The file was not changed.`);
+    }
+    if (missions.enabled !== undefined &&
+        config.fleetView !== undefined &&
+        config.asyncByDefault !== undefined &&
+        config.singleRunOutputBaseDir !== undefined) {
         return undefined;
     }
     const next = {
@@ -103,6 +109,10 @@ function prepareSubagentDefaults(settingsPath) {
         missions: { ...missions, enabled: missions.enabled ?? false },
         fleetView: config.fleetView ?? false,
         asyncByDefault: config.asyncByDefault ?? true,
+        // pi-subagents otherwise saves a relative `output` such as
+        // outputs/.drafts/x.md under the session's subagent-artifacts folder, where
+        // the lead agent never looks. "." resolves against Pi's cwd, the workspace.
+        singleRunOutputBaseDir: config.singleRunOutputBaseDir ?? ".",
     };
     return { path, original: existing?.source, content: `${JSON.stringify(next, null, 2)}\n` };
 }

package/extensions/research-tools/request-pacer.ts

@@ -0,0 +1,17 @@
+// Runs requests one after another, each starting at least `minGapMs` after the
+// previous one started. Pi runs parallel tool calls in one process, so this keeps
+// them under a service's per-second limit.
+export function createRequestPacer(): <T>(minGapMs: number, run: () => Promise<T>) => Promise<T> {
+	let queue: Promise<unknown> = Promise.resolve();
+	let lastStartedAt = Number.NEGATIVE_INFINITY;
+	return <T>(minGapMs: number, run: () => Promise<T>): Promise<T> => {
+		const next = queue.then(async () => {
+			const wait = lastStartedAt + minGapMs - Date.now();
+			if (wait > 0) await new Promise((resolve) => setTimeout(resolve, wait));
+			lastStartedAt = Date.now();
+			return run();
+		});
+		queue = next.catch(() => {});
+		return next;
+	};
+}

package/extensions/research-tools/science-database-openalex.ts

@@ -1,4 +1,5 @@
 import { OPENALEX_API_KEY_HINT, openAlexRequestFailure, searchExactOpenAlex } from "./science-database-openalex-exact.js";
+import { createRequestPacer } from "./request-pacer.js";
 
 export type OpenAlexScienceDatabaseSource = "openalex";
 
@@ -96,9 +97,18 @@ function scrubOpenAlexText(text: string, url: URL): string {
 	return scrubbed;
 }
 
-async function fetchJson(url: URL): Promise<{ credentialStatus: string; endpoint: string; payload: unknown; usingApiKey: boolean }> {
-	const auth = addAuth(url);
-	const endpoint = scrubOpenAlexEndpoint(url);
+type OpenAlexResponse =
+	| { ok: true; payload: unknown }
+	| { ok: false; status: number; statusText: string; retryAfterMs?: number; body: string };
+
+// Semantic search allows one request per second (429 with Retry-After: 1), so
+// semantic requests are paced. A 429 that asks for a short wait, such as one
+// caused by a subagent in another process, is retried once.
+const semanticPacer = createRequestPacer();
+const SEMANTIC_MIN_GAP_MS = 1100;
+const MAX_RETRY_WAIT_MS = 5000;
+
+async function requestOnce(url: URL): Promise<OpenAlexResponse> {
 	const controller = new AbortController();
 	const timeout = setTimeout(() => controller.abort(), REQUEST_TIMEOUT_MS);
 	try {
@@ -109,16 +119,39 @@ async function fetchJson(url: URL): Promise<{ credentialStatus: string; endpoint
 			},
 			signal: controller.signal,
 		});
-		if (!response.ok) {
-			const snippet = scrubOpenAlexText((await response.text()).slice(0, 4096), url).slice(0, 240);
-			throw openAlexRequestFailure(response.status, response.statusText, snippet, auth.usingApiKey);
-		}
-		return { ...auth, endpoint, payload: await response.json() };
+		if (response.ok) return { ok: true, payload: await response.json() };
+		const retryAfter = response.headers.get("retry-after");
+		const retryAfterSeconds = retryAfter === null ? Number.NaN : Number(retryAfter);
+		return {
+			ok: false,
+			status: response.status,
+			statusText: response.statusText,
+			retryAfterMs: Number.isFinite(retryAfterSeconds) && retryAfterSeconds >= 0 ? retryAfterSeconds * 1000 : undefined,
+			body: (await response.text()).slice(0, 4096),
+		};
 	} finally {
 		clearTimeout(timeout);
 	}
 }
 
+async function fetchJson(url: URL): Promise<{ credentialStatus: string; endpoint: string; payload: unknown; usingApiKey: boolean }> {
+	const auth = addAuth(url);
+	const endpoint = scrubOpenAlexEndpoint(url);
+	const request = () =>
+		url.searchParams.has("search.semantic") ? semanticPacer(SEMANTIC_MIN_GAP_MS, () => requestOnce(url)) : requestOnce(url);
+	let response = await request();
+	if (!response.ok && response.status === 429 && response.retryAfterMs !== undefined && response.retryAfterMs <= MAX_RETRY_WAIT_MS) {
+		const waitMs = response.retryAfterMs + Math.random() * 500;
+		await new Promise((resolve) => setTimeout(resolve, waitMs));
+		response = await request();
+	}
+	if (!response.ok) {
+		const snippet = scrubOpenAlexText(response.body, url).slice(0, 240);
+		throw openAlexRequestFailure(response.status, response.statusText, snippet, auth.usingApiKey);
+	}
+	return { ...auth, endpoint, payload: response.payload };
+}
+
 function shortOpenAlexId(value: unknown): string | undefined {
 	const raw = stringValue(value);
 	if (!raw) return undefined;

package/extensions/research-tools/science-databases.ts

@@ -5,6 +5,7 @@ import { searchOpenAlex } from "./science-database-openalex.js";
 import { searchPreprints } from "./science-database-preprints.js";
 import { searchPubMed } from "./science-database-pubmed.js";
 import { isEuropePmcFullTextQuery, searchEuropePmcFullText } from "./science-database-europepmc-fulltext.js";
+import { createRequestPacer } from "./request-pacer.js";
 
 type ScienceDatabaseSource = "arxiv" | "biorxiv" | "crossref" | "europepmc" | "medrxiv" | "openalex" | "pubmed" | "semanticscholar";
 
@@ -141,21 +142,12 @@ function crossrefAuthors(value: unknown): string[] {
 // (x-concurrency-limit: 1); the polite pool, used when a mailto is sent, allows
 // three per second. Parallel DOI checks got 429 on most calls, so run Crossref
 // requests one after another at the pool's pace.
-let crossrefQueue: Promise<unknown> = Promise.resolve();
-let crossrefLastStartedAt = Number.NEGATIVE_INFINITY;
+const crossrefPacer = createRequestPacer();
 
 export function withCrossrefPacing<T>(polite: boolean, run: () => Promise<T>): Promise<T> {
 	// A little over the pool's limit: network jitter can land two requests paced
 	// exactly one second apart inside the same window.
-	const minGapMs = polite ? 400 : 1200;
-	const next = crossrefQueue.then(async () => {
-		const wait = crossrefLastStartedAt + minGapMs - Date.now();
-		if (wait > 0) await new Promise((resolve) => setTimeout(resolve, wait));
-		crossrefLastStartedAt = Date.now();
-		return run();
-	});
-	crossrefQueue = next.catch(() => {});
-	return next;
+	return crossrefPacer(polite ? 400 : 1200, run);
 }
 
 async function searchCrossref(params: ScienceDatabaseSearchParams): Promise<Record<string, unknown>> {

package/metadata/commands.mjs

@@ -115,7 +115,7 @@ export const cliCommandSections = [
 			{ usage: "feynman model list", description: "List available models in Pi auth storage." },
 			{ usage: "feynman model login [id]", description: "Authenticate a model provider with OAuth or API-key setup." },
 			{ usage: "feynman model logout [id]", description: "Clear stored auth for a model provider." },
-			{ usage: "feynman model set <provider/model>", description: "Set the default approved research model (also accepts provider:model)." },
+			{ usage: "feynman model set [provider/model]", description: "Set the default approved research model (also accepts provider:model); without one, pick from a list." },
 			{ usage: "feynman model tier [value]", description: "View or set the request service tier override." },
 		],
 	},
@@ -137,6 +137,7 @@ export const cliCommandSections = [
 		commands: [
 			{ usage: "feynman packages list", description: "Show core and optional Pi package presets." },
 			{ usage: "feynman packages install <preset>", description: "Install optional package presets on demand." },
+			{ usage: "feynman packages remove <preset>", description: "Remove an installed optional package preset." },
 			{ usage: "feynman search status", description: "Show Pi web-access status and config path." },
 			{ usage: "feynman search set <provider> [api-key]", description: "Set the web search provider and optionally save its API key." },
 			{ usage: "feynman search clear", description: "Reset web search provider to auto while preserving API keys." },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@companion-ai/feynman",
-  "version": "0.5.14",
+  "version": "0.5.16",
   "description": "Research-first CLI agent built on Pi and alphaXiv",
   "license": "MIT",
   "type": "module",
@@ -79,18 +79,17 @@
   "dependencies": {
     "@clack/prompts": "^1.8.1",
     "@companion-ai/alpha-hub": "0.1.6",
-    "@earendil-works/pi-ai": "*",
-    "@earendil-works/pi-coding-agent": "*",
     "fast-xml-parser": "5.11.1",
     "pi-btw": "0.6.1",
     "pi-docparser": "4.0.0",
     "pi-subagents": "0.73.1",
-    "pi-web-access": "0.33.0",
-    "posthog-node": "^5.53.0"
+    "pi-web-access": "0.34.0",
+    "posthog-node": "^5.53.0",
+    "undici": "^8.11.2"
   },
   "peerDependencies": {
-    "@earendil-works/pi-ai": "*",
-    "@earendil-works/pi-coding-agent": "*",
+    "@earendil-works/pi-ai": "0.99.1",
+    "@earendil-works/pi-coding-agent": "0.99.1",
     "@earendil-works/pi-tui": "*",
     "typebox": "*"
   },
@@ -117,6 +116,6 @@
     "access": "public"
   },
   "overrides": {
-    "undici": "^8.10.2"
+    "undici": "$undici"
   }
 }

package/prompts/deepresearch.md

@@ -76,13 +76,13 @@ Example shape:
 
 ```json
 {
-  "workflowScript": "return await runs.all([{key:'web',agent:'researcher',task:'Read outputs/.plans/<slug>-T1.md and write <slug>-research-web.md.',output:'<slug>-research-web.md'},{key:'papers',agent:'researcher',task:'Read outputs/.plans/<slug>-T2.md and write <slug>-research-papers.md.',output:'<slug>-research-papers.md'}]);",
+  "workflowScript": "return await runs.all([{key:'web',agent:'researcher',task:'Read outputs/.plans/<slug>-T1.md and write outputs/.drafts/<slug>-research-web.md.',output:'outputs/.drafts/<slug>-research-web.md'},{key:'papers',agent:'researcher',task:'Read outputs/.plans/<slug>-T2.md and write outputs/.drafts/<slug>-research-papers.md.',output:'outputs/.drafts/<slug>-research-papers.md'}]);",
   "async": true,
   "globalConcurrencyLimit": 4
 }
 ```
 
-Continue independent work after launch, then consume completion results before synthesis. Use the returned output references to locate managed child files; verify them on disk and copy them to the planned research paths when necessary. After evidence gathering, update the plan ledger and verification log. If research failed, record exactly what failed and proceed with a blocked or partial draft.
+Continue independent work after launch, then consume completion results before synthesis. Relative output paths resolve against the workspace, so each child writes its declared path; verify each file on disk, and use the returned output reference for any that is missing. After evidence gathering, update the plan ledger and verification log. If research failed, record exactly what failed and proceed with a blocked or partial draft.
 
 ## Step 4: Draft
 
@@ -123,7 +123,7 @@ Use this shape:
 }
 ```
 
-Wait for the verifier's completion result before review, not merely the async launch receipt. Verify on disk that `outputs/.drafts/<slug>-cited.md` exists. If managed output routing wrote elsewhere, use the returned output reference to find the cited file and move or copy it to `outputs/.drafts/<slug>-cited.md`.
+Wait for the verifier's completion result before review, not merely the async launch receipt. Verify on disk that `outputs/.drafts/<slug>-cited.md` exists. If it is missing, use the returned output reference to find the cited file and copy it there.
 
 ## Step 6: Review
 
@@ -142,7 +142,7 @@ Use this shape:
   "agent": "reviewer",
   "async": true,
   "task": "Verify outputs/.drafts/<slug>-cited.md. Flag unsupported claims, logical gaps, single-source critical claims, and overstated confidence. This is a verification pass, not a peer review.",
-  "output": "<slug>-verification.md"
+  "output": "outputs/.drafts/<slug>-verification.md"
 }
 ```
 

package/prompts/lit.md

@@ -12,7 +12,7 @@ Derive a short slug from the topic (lowercase, hyphens, no filler words, ≤5 wo
 
 1. **Plan** — Outline the scope: key questions, source types to search (papers, web, repos), time period, expected sections, and a small task ledger plus verification log. When the input appears to name a lab, PI, author, institution lab page, or author profile, run the review as a publication-corpus review: find the lab/author identity first, collect the reachable publication list, then map the research trajectory across that corpus. Write the plan to `outputs/.plans/<slug>.md`. Briefly summarize the plan to the user 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. Do not ask for confirmation or wait for a proceed response unless the user explicitly requested plan review.
    - When updating the plan ledger later, keep edits small and valid. If an `edit` tool call fails with a JSON parse error or the replacement would require embedding a large markdown block, rewrite the full corrected plan file with the file-writing tool instead, then continue to final artifact/provenance verification.
-2. **Gather** — Use the `researcher` subagent when the sweep is wide enough to benefit from delegated paper triage before synthesis. For narrow topics, search directly. Researcher outputs go to `<slug>-research-*.md`. For publication-corpus reviews, the lead agent owns identity resolution and writes `notes/<slug>-publications.md` with reachable titles, years, venues, URLs/DOIs, and gaps before delegating trajectory synthesis. Prefer lab publication pages, author profiles, arXiv/OpenReview/Semantic Scholar pages, and paper search results that expose stable source URLs. Do not silently skip assigned questions; mark them `done`, `blocked`, or `superseded`.
+2. **Gather** — Use the `researcher` subagent when the sweep is wide enough to benefit from delegated paper triage before synthesis. For narrow topics, search directly. Researcher outputs go to `outputs/.drafts/<slug>-research-*.md`; pass that path as each researcher's `output`. For publication-corpus reviews, the lead agent owns identity resolution and writes `notes/<slug>-publications.md` with reachable titles, years, venues, URLs/DOIs, and gaps before delegating trajectory synthesis. Prefer lab publication pages, author profiles, arXiv/OpenReview/Semantic Scholar pages, and paper search results that expose stable source URLs. Do not silently skip assigned questions; mark them `done`, `blocked`, or `superseded`.
 3. **Synthesize** — Separate consensus, disagreements, and open questions. For publication-corpus reviews, also identify 3-5 research trajectories and the 3-5 papers that most changed the corpus direction; rank them by contrastive originality, methodology strength, and relationship to prior art rather than by author prestige alone. When useful, propose concrete next experiments or follow-up reading. Use Mermaid diagrams for taxonomies, method pipelines, or lab trajectory maps when the structure is source-supported and changes the reader's research decision. Keep the output to research evidence, source coverage, and next research decisions; do not create non-research operational artifacts from a literature review run.
 4. **Cite** — Spawn the `verifier` agent to add inline citations and verify every source URL in the draft.
 5. **Verify** — Spawn the `reviewer` agent to check the cited draft for unsupported claims, logical gaps, zombie sections, and single-source critical findings. Fix FATAL issues before delivering. Note MAJOR issues in Open Questions. If FATAL issues were found, run one more verification pass after the fixes.

package/prompts/summarize.md

@@ -126,7 +126,7 @@ Use one async workflow with a stable key for each chunk and a concurrency limit
 }
 ```
 
-Consume completion results before aggregation. `runs.all` returns an ordered array including ordinary child failures; inspect each `ok` and error. A validation or infrastructure failure can still fail the workflow. Use actual returned output/artifact references to locate child files, verify them on disk, and copy them to the expected chunk-summary paths when necessary. Record missing or failed chunks rather than treating an async launch receipt as a finished summary.
+Consume completion results before aggregation. `runs.all` returns an ordered array including ordinary child failures; inspect each `ok` and error. A validation or infrastructure failure can still fail the workflow. Relative output paths resolve against the workspace; verify each chunk summary on disk, and use the returned output reference for any that is missing. Record missing or failed chunks rather than treating an async launch receipt as a finished summary.
 
 ### 3d. Aggregate