@openhoo/hoopilot 2.1.3 → 2.1.5

package/README.md

@@ -66,7 +66,7 @@ $env:OPENAI_BASE_URL = "http://127.0.0.1:4141/v1"
 $env:OPENAI_API_KEY = "hoopilot"
 ```
 
-To require clients to authenticate — recommended whenever you expose the proxy beyond localhost — set `HOOPILOT_API_KEY` to a strong, unique secret and send that value as the client key:
+To require clients to authenticate — recommended whenever you expose the proxy beyond localhost — set `HOOPILOT_API_KEY` to a strong, unique secret of at least 24 characters and send that value as the client key:
 
 ```sh
 export HOOPILOT_API_KEY=$(openssl rand -hex 24)
@@ -161,7 +161,7 @@ Tags follow the release version, for example `ghcr.io/openhoo/hoopilot:1.3`, `:1
 
 #### Exposing the proxy beyond loopback
 
-The image binds `0.0.0.0` and cannot tell whether the published port is loopback-only, so it fails closed: drop the `-e HOOPILOT_ALLOW_UNAUTHENTICATED=1` opt-in (or map the port to a non-loopback interface) and it refuses to start without a strong, unique `HOOPILOT_API_KEY` (well-known demo keys are rejected). Clients then send that key as `Authorization: Bearer <key>` or `x-api-key: <key>`:
+The image binds `0.0.0.0` and cannot tell whether the published port is loopback-only, so it fails closed: drop the `-e HOOPILOT_ALLOW_UNAUTHENTICATED=1` opt-in (or map the port to a non-loopback interface) and it refuses to start without a strong, unique `HOOPILOT_API_KEY` of at least 24 characters. Short, repeated, and well-known demo keys are rejected. Clients then send that key as `Authorization: Bearer <key>` or `x-api-key: <key>`:
 
 ```sh
 export HOOPILOT_API_KEY=$(openssl rand -hex 24)
@@ -201,7 +201,7 @@ Start the server:
 hoopilot --port 4141
 ```
 
-By default Hoopilot listens on `127.0.0.1:4141`. If `HOOPILOT_API_KEY` is unset, local requests are accepted without client authentication. Binding to a non-loopback host requires either a strong, unique `HOOPILOT_API_KEY` or the explicit `--allow-unauthenticated` / `HOOPILOT_ALLOW_UNAUTHENTICATED=1` opt-in. Well-known demo keys are always rejected on a non-loopback host, even with the unauthenticated opt-in.
+By default Hoopilot listens on `127.0.0.1:4141`. If `HOOPILOT_API_KEY` is unset, local requests are accepted without client authentication. Binding to a non-loopback host requires either a strong, unique `HOOPILOT_API_KEY` of at least 24 characters or the explicit `--allow-unauthenticated` / `HOOPILOT_ALLOW_UNAUTHENTICATED=1` opt-in. Short, repeated, and well-known demo keys are always rejected on a non-loopback host, even with the unauthenticated opt-in.
 
 When an API key is configured, clients may send it as either `Authorization: Bearer <key>` or `x-api-key: <key>`.
 
@@ -250,6 +250,8 @@ With the [keyless Docker quick start](#keyless-local-quick-start), no key is inv
 
 `codexx` does not start Hoopilot and does not alter your shell environment. It starts `codex` with a temporary `hoopilot` model provider pointed at `http://127.0.0.1:4141/v1`, uses the Responses API wire format, disables Responses WebSockets for that provider, maps `HOOPILOT_API_KEY` (or a random throwaway key when none is set) to `OPENAI_API_KEY` for the child process, disables Codex's managed network proxy (`-c permissions.workspace.network.enabled=false`), and removes standard proxy variables from the spawned Codex process.
 
+`codexx` also sets Codex's temporary provider stream idle timeout to 120 seconds by default, so a silent Responses stream fails and Codex's normal stream retry path can run instead of leaving the turn apparently stuck. Override with `CODEXX_STREAM_IDLE_TIMEOUT_MS`, or set it to `0` to use Codex's own default.
+
 > **Why the network proxy is disabled.** Codex ships a managed network proxy that routes the agent's traffic through a local proxy (port `3128`) and enforces a domain allowlist. Because the local Hoopilot server isn't on that allowlist, leaving the proxy on makes Codex's request to Hoopilot fail with an instant `403` Squid error. The proxy has two independent gates — the `network_proxy` feature flag and the `permissions.workspace.network.enabled` config — and disabling only the feature is not always enough, so `codexx` turns off both (`--disable network_proxy` and `-c permissions.workspace.network.enabled=false`) and requests reach Hoopilot directly.
 
 `codexx` defaults to `gpt-5.5` with `model_reasoning_effort="xhigh"`. Before starting Codex, it checks `/v1/models` and reports if the logged-in Copilot account does not advertise the requested model. Set `CODEXX_MODEL` to one of the listed models, or log in with a Copilot account that has access to the default model.
@@ -364,7 +366,7 @@ Server and local-client settings:
 | --- | --- |
 | `HOST` / `--host` | Host to listen on. Default: `127.0.0.1` for local runs; Docker sets `0.0.0.0`. |
 | `PORT` / `--port` | Port to listen on. Default: `4141`. |
-| `HOOPILOT_API_KEY` / `--api-key` | Require clients to send `Authorization: Bearer <key>` or `x-api-key: <key>`. Must be a strong, unique secret on non-loopback binds; well-known demo keys are rejected. |
+| `HOOPILOT_API_KEY` / `--api-key` | Require clients to send `Authorization: Bearer <key>` or `x-api-key: <key>`. Must be a strong, unique secret of at least 24 characters on non-loopback binds; short, repeated, and well-known demo keys are rejected. |
 | `--api-key-file` | Read the local API key from a file instead of argv. |
 | `HOOPILOT_ALLOWED_ORIGINS` | Comma-separated browser origins allowed to make cross-origin requests. Loopback origins are always allowed; every other origin is blocked. |
 | `HOOPILOT_ALLOW_UNAUTHENTICATED` / `--allow-unauthenticated` | Allow non-loopback binds without a local API key. |
@@ -387,6 +389,8 @@ Logging and update settings:
 | --- | --- |
 | `HOOPILOT_LOG_LEVEL` / `--log-level` | `trace`, `debug`, `info`, `warn`, `error`, `fatal`, or `silent`. Default: `info`. |
 | `HOOPILOT_LOG_FORMAT` / `--log-format` | `pretty` or `json`. Default: `pretty`. |
+| `HOOPILOT_UPSTREAM_TIMEOUT_MS` | Time to wait for Copilot response headers before returning `504 copilot_timeout`. Default: `120000`; set `0` to disable. |
+| `HOOPILOT_UPSTREAM_STREAM_IDLE_TIMEOUT_MS` | Time to wait for bytes on a Copilot response body before failing the stream so clients can retry. Default: `120000`; set `0` to disable. |
 | `HOOPILOT_NO_UPDATE_CHECK` / `--no-update-check` | Disable background update checks. `NO_UPDATE_NOTIFIER` is also honored. |
 
 `codexx` settings:
@@ -398,6 +402,7 @@ Logging and update settings:
 | `CODEXX_CODEX_BIN` | Codex executable to run. Default: `codex`. |
 | `CODEXX_MODEL` | Codex model to use. Default: `gpt-5.5`. |
 | `CODEXX_MODEL_REASONING_EFFORT` | Codex reasoning effort. Default: `xhigh`. |
+| `CODEXX_STREAM_IDLE_TIMEOUT_MS` | Codex Responses stream idle timeout for the temporary Hoopilot provider. Default: `120000`; set `0` to use Codex's own default. |
 | `CODEXX_SKIP_MODEL_PREFLIGHT=1` | Skip the `/v1/models` availability check before starting Codex. |
 
 ## CLI reference

package/dist/{chunk-CYR6I4C3.js → chunk-4ZG5QEYJ.js}

@@ -116,6 +116,7 @@ var DEFAULT_BASE_URL = "http://127.0.0.1:4141/v1";
 var DEFAULT_CODEX_BIN = "codex";
 var DEFAULT_MODEL = "gpt-5.5";
 var DEFAULT_REASONING_EFFORT = "xhigh";
+var DEFAULT_STREAM_IDLE_TIMEOUT_MS = 12e4;
 var PROXY_ENV_KEYS = [
   "ALL_PROXY",
   "HTTPS_PROXY",
@@ -132,13 +133,18 @@ function buildCodexxInvocation(argv, env = process.env) {
   const command = envValue(env.CODEXX_CODEX_BIN) ?? DEFAULT_CODEX_BIN;
   const model = envValue(env.CODEXX_MODEL) ?? DEFAULT_MODEL;
   const reasoningEffort = envValue(env.CODEXX_MODEL_REASONING_EFFORT) ?? DEFAULT_REASONING_EFFORT;
-  const providerConfig = [
+  const streamIdleTimeoutMs = parseStreamIdleTimeoutMs(env.CODEXX_STREAM_IDLE_TIMEOUT_MS);
+  const providerConfigParts = [
     '{ name = "Hoopilot"',
     `base_url = ${JSON.stringify(baseUrl)}`,
     'env_key = "OPENAI_API_KEY"',
     'wire_api = "responses"',
-    "supports_websockets = false }"
-  ].join(", ");
+    "supports_websockets = false"
+  ];
+  if (streamIdleTimeoutMs > 0) {
+    providerConfigParts.push(`stream_idle_timeout_ms = ${streamIdleTimeoutMs}`);
+  }
+  const providerConfig = `${providerConfigParts.join(", ")} }`;
   return {
     args: [
       // Codex ships a managed network proxy (codex-rs/network-proxy) that routes the
@@ -176,6 +182,17 @@ function buildCodexxInvocation(argv, env = process.env) {
 function generateEphemeralApiKey() {
   return `codexx-${crypto.randomUUID()}`;
 }
+function parseStreamIdleTimeoutMs(rawValue) {
+  const raw = envValue(rawValue);
+  if (raw === void 0) {
+    return DEFAULT_STREAM_IDLE_TIMEOUT_MS;
+  }
+  const value = Number(raw);
+  if (!Number.isInteger(value) || value < 0) {
+    throw new Error("CODEXX_STREAM_IDLE_TIMEOUT_MS must be a non-negative integer.");
+  }
+  return value;
+}
 function withoutProxyEnv(env) {
   const next = { ...env };
   for (const key of PROXY_ENV_KEYS) {
@@ -261,6 +278,9 @@ Environment:
   CODEXX_MODEL         Codex model to use. Default: ${DEFAULT_MODEL}
   CODEXX_MODEL_REASONING_EFFORT
                        Codex reasoning effort. Default: ${DEFAULT_REASONING_EFFORT}
+  CODEXX_STREAM_IDLE_TIMEOUT_MS
+                       Codex Responses stream idle timeout in milliseconds. Default:
+                       ${DEFAULT_STREAM_IDLE_TIMEOUT_MS}; set 0 to use Codex's own default.
   CODEXX_SKIP_MODEL_PREFLIGHT
                        Set to 1 to skip checking /v1/models before starting Codex.
 
@@ -295,4 +315,4 @@ export {
   main,
   verifyCodexxModel
 };
-//# sourceMappingURL=chunk-CYR6I4C3.js.map
+//# sourceMappingURL=chunk-4ZG5QEYJ.js.map

package/dist/chunk-4ZG5QEYJ.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/codexx.ts","../src/util.ts"],"sourcesContent":["#!/usr/bin/env bun\n\nimport { spawn } from \"node:child_process\";\nimport { constants as osConstants } from \"node:os\";\nimport type { FetchLike } from \"./types\";\nimport {\n  envValue,\n  errorMessage,\n  modelIdsFromResponse,\n  trimTrailingSlash,\n  truncatedResponseText,\n} from \"./util\";\n\nconst DEFAULT_BASE_URL = \"http://127.0.0.1:4141/v1\";\nconst DEFAULT_CODEX_BIN = \"codex\";\nconst DEFAULT_MODEL = \"gpt-5.5\";\nconst DEFAULT_REASONING_EFFORT = \"xhigh\";\nconst DEFAULT_STREAM_IDLE_TIMEOUT_MS = 120_000;\nconst PROXY_ENV_KEYS = [\n  \"ALL_PROXY\",\n  \"HTTPS_PROXY\",\n  \"HTTP_PROXY\",\n  \"NO_PROXY\",\n  \"all_proxy\",\n  \"https_proxy\",\n  \"http_proxy\",\n  \"no_proxy\",\n];\n\nexport interface CodexxInvocation {\n  args: string[];\n  baseUrl: string;\n  command: string;\n  env: NodeJS.ProcessEnv;\n  model: string;\n}\n\nexport function buildCodexxInvocation(\n  argv: string[],\n  env: NodeJS.ProcessEnv = process.env,\n): CodexxInvocation {\n  const baseUrl = envValue(env.CODEXX_BASE_URL) ?? DEFAULT_BASE_URL;\n  // Never fall back to a public, predictable key: a shared constant like the old\n  // \"local-key\" default is also a credential a malicious local/browser client\n  // could guess. When no key is configured the local server is expected to run\n  // unauthenticated, which accepts any value, so a random throwaway key is safe.\n  const apiKey =\n    envValue(env.CODEXX_API_KEY) ?? envValue(env.HOOPILOT_API_KEY) ?? generateEphemeralApiKey();\n  const command = envValue(env.CODEXX_CODEX_BIN) ?? DEFAULT_CODEX_BIN;\n  const model = envValue(env.CODEXX_MODEL) ?? DEFAULT_MODEL;\n  const reasoningEffort = envValue(env.CODEXX_MODEL_REASONING_EFFORT) ?? DEFAULT_REASONING_EFFORT;\n  const streamIdleTimeoutMs = parseStreamIdleTimeoutMs(env.CODEXX_STREAM_IDLE_TIMEOUT_MS);\n  const providerConfigParts = [\n    '{ name = \"Hoopilot\"',\n    `base_url = ${JSON.stringify(baseUrl)}`,\n    'env_key = \"OPENAI_API_KEY\"',\n    'wire_api = \"responses\"',\n    \"supports_websockets = false\",\n  ];\n  if (streamIdleTimeoutMs > 0) {\n    providerConfigParts.push(`stream_idle_timeout_ms = ${streamIdleTimeoutMs}`);\n  }\n  const providerConfig = `${providerConfigParts.join(\", \")} }`;\n\n  return {\n    args: [\n      // Codex ships a managed network proxy (codex-rs/network-proxy) that routes the\n      // agent's traffic through a local proxy on :3128 and enforces a domain allowlist.\n      // A host that is not allowlisted — like the local Hoopilot server — gets an instant\n      // 403 (the Squid error page) and never reaches Hoopilot. It has two independent\n      // gates: the `network_proxy` feature flag and the `permissions.workspace.network`\n      // config. Disabling only the feature (`--disable network_proxy`) does not reliably\n      // turn it off when the proxy is enabled through the permissions config, so set both\n      // off: the feature flag via `--disable`, and the proxy itself via the config key\n      // (when `enabled` is false the proxy no-ops and binds no listeners).\n      \"--disable\",\n      \"network_proxy\",\n      \"-c\",\n      \"permissions.workspace.network.enabled=false\",\n      \"-c\",\n      'model_provider=\"hoopilot\"',\n      \"-c\",\n      `model_providers.hoopilot=${providerConfig}`,\n      \"-m\",\n      model,\n      \"-c\",\n      `model_reasoning_effort=${JSON.stringify(reasoningEffort)}`,\n      ...argv,\n    ],\n    baseUrl,\n    command,\n    env: withoutProxyEnv({\n      ...env,\n      OPENAI_API_KEY: apiKey,\n    }),\n    model,\n  };\n}\n\n// A random, non-guessable placeholder key for when neither CODEXX_API_KEY nor\n// HOOPILOT_API_KEY is set. An unauthenticated local Hoopilot accepts any value;\n// a keyed server rejects it with a 401, which the model preflight surfaces.\nfunction generateEphemeralApiKey(): string {\n  return `codexx-${crypto.randomUUID()}`;\n}\n\nfunction parseStreamIdleTimeoutMs(rawValue: string | undefined): number {\n  const raw = envValue(rawValue);\n  if (raw === undefined) {\n    return DEFAULT_STREAM_IDLE_TIMEOUT_MS;\n  }\n  const value = Number(raw);\n  if (!Number.isInteger(value) || value < 0) {\n    throw new Error(\"CODEXX_STREAM_IDLE_TIMEOUT_MS must be a non-negative integer.\");\n  }\n  return value;\n}\n\nfunction withoutProxyEnv(env: NodeJS.ProcessEnv): NodeJS.ProcessEnv {\n  const next = { ...env };\n  for (const key of PROXY_ENV_KEYS) {\n    delete next[key];\n  }\n  return next;\n}\n\nexport async function main(argv = Bun.argv.slice(2), env = process.env): Promise<void> {\n  if (argv.length === 1 && (argv[0] === \"--help\" || argv[0] === \"-h\")) {\n    console.log(helpText());\n    return;\n  }\n\n  const invocation = buildCodexxInvocation(argv, env);\n  if (env.CODEXX_SKIP_MODEL_PREFLIGHT !== \"1\") {\n    await verifyCodexxModel(invocation);\n  }\n  const child = spawn(invocation.command, invocation.args, {\n    env: invocation.env,\n    shell: process.platform === \"win32\",\n    stdio: \"inherit\",\n  });\n\n  const exitCode = await new Promise<number>((resolve, reject) => {\n    child.once(\"error\", reject);\n    child.once(\"exit\", (code, signal) => {\n      if (typeof code === \"number\") {\n        resolve(code);\n        return;\n      }\n      resolve(signal ? 128 + signalNumber(signal) : 1);\n    });\n  });\n\n  process.exitCode = exitCode;\n}\n\nexport async function verifyCodexxModel(\n  invocation: Pick<CodexxInvocation, \"baseUrl\" | \"env\" | \"model\">,\n  fetcher: FetchLike = fetch,\n): Promise<void> {\n  const modelsUrl = `${trimTrailingSlash(invocation.baseUrl)}/models`;\n  const apiKey = invocation.env.OPENAI_API_KEY;\n  if (apiKey === undefined) {\n    throw new Error(\n      \"verifyCodexxModel requires invocation.env.OPENAI_API_KEY; build the invocation with buildCodexxInvocation.\",\n    );\n  }\n  let response: Response;\n  try {\n    response = await fetcher(modelsUrl, {\n      headers: {\n        accept: \"application/json\",\n        authorization: `Bearer ${apiKey}`,\n      },\n      method: \"GET\",\n    });\n  } catch (error) {\n    throw new Error(\n      `Could not reach Hoopilot at ${modelsUrl}. Start Hoopilot first, or set CODEXX_SKIP_MODEL_PREFLIGHT=1 to skip this check. ${errorMessage(error)}`,\n    );\n  }\n\n  if (!response.ok) {\n    throw new Error(\n      `Could not verify model ${JSON.stringify(invocation.model)} because ${modelsUrl} returned ${response.status}: ${await truncatedResponseText(response)}`,\n    );\n  }\n\n  const models = modelIdsFromResponse(await response.json().catch(() => undefined));\n  if (models.length > 0 && !models.includes(invocation.model)) {\n    throw new Error(\n      `The logged-in Copilot account does not advertise model ${JSON.stringify(invocation.model)} at ${modelsUrl}. Available models: ${models.join(\", \")}. After upgrading Hoopilot, rerun \"hoopilot login\" to refresh the Copilot OAuth token, or set CODEXX_MODEL to one of the advertised model IDs.`,\n    );\n  }\n}\n\nfunction helpText(): string {\n  return `codexx\n\nRun Codex against an already-running local Hoopilot server.\n\nUsage:\n  codexx [codex options] [prompt]\n\nEnvironment:\n  CODEXX_BASE_URL      OpenAI-compatible base URL. Default: ${DEFAULT_BASE_URL}\n  CODEXX_API_KEY       API key sent to the local Hoopilot server.\n  HOOPILOT_API_KEY     Used as the API key when CODEXX_API_KEY is unset. When\n                       neither is set, a random throwaway key is generated for\n                       an unauthenticated local server.\n  CODEXX_CODEX_BIN     Codex executable to run. Default: ${DEFAULT_CODEX_BIN}\n  CODEXX_MODEL         Codex model to use. Default: ${DEFAULT_MODEL}\n  CODEXX_MODEL_REASONING_EFFORT\n                       Codex reasoning effort. Default: ${DEFAULT_REASONING_EFFORT}\n  CODEXX_STREAM_IDLE_TIMEOUT_MS\n                       Codex Responses stream idle timeout in milliseconds. Default:\n                       ${DEFAULT_STREAM_IDLE_TIMEOUT_MS}; set 0 to use Codex's own default.\n  CODEXX_SKIP_MODEL_PREFLIGHT\n                       Set to 1 to skip checking /v1/models before starting Codex.\n\ncodexx does not start Hoopilot and does not change your shell environment. It selects a temporary Hoopilot model provider with Responses WebSockets disabled, uses ${DEFAULT_MODEL} with ${DEFAULT_REASONING_EFFORT} reasoning by default, disables Codex's managed network proxy (permissions.workspace.network.enabled=false) so requests reach the local server instead of being blocked by its allowlist, and removes proxy variables only from the spawned Codex process.`;\n}\n\nfunction signalNumber(signal: NodeJS.Signals): number {\n  return osConstants.signals[signal] ?? 1;\n}\n\nif (import.meta.main) {\n  main().catch((error: unknown) => {\n    console.error(errorMessage(error));\n    process.exit(1);\n  });\n}\n","import type { JsonObject, StreamingProxyMode } from \"./types\";\n\n/** Remove any trailing slashes from a URL or path string. */\nexport function trimTrailingSlash(value: string): string {\n  return value.replace(/\\/+$/, \"\");\n}\n\n/** Treat blank environment variables as unset while preserving nonblank values. */\nexport function envValue(value: string | undefined): string | undefined {\n  const trimmed = value?.trim();\n  return trimmed ? trimmed : undefined;\n}\n\n/** True for HTTPS URLs, or HTTP only on loopback hosts used by local tests/dev. */\nexport function isHttpsOrLoopbackUrl(rawUrl: string): boolean {\n  const url = parseUrl(rawUrl);\n  if (!url) {\n    return false;\n  }\n  return url.protocol === \"https:\" || isLoopbackHttpUrl(url);\n}\n\n/** Validate a base URL before sending a bearer/OAuth token to it. */\nexport function isTrustedTokenBaseUrl(\n  rawUrl: string,\n  allowedHttpsHosts: readonly string[],\n  allowUnsafeHttps = false,\n): boolean {\n  const url = parseUrl(rawUrl);\n  if (!url) {\n    return false;\n  }\n  if (url.username || url.password || url.search || url.hash) {\n    return false;\n  }\n  if (url.pathname !== \"\" && url.pathname !== \"/\") {\n    return false;\n  }\n  if (isLoopbackHttpUrl(url)) {\n    return true;\n  }\n  if (url.protocol !== \"https:\") {\n    return false;\n  }\n  const host = url.hostname.toLowerCase();\n  return allowedHttpsHosts.includes(host) || allowUnsafeHttps;\n}\n\nfunction parseUrl(rawUrl: string): URL | undefined {\n  let url: URL;\n  try {\n    url = new URL(rawUrl);\n  } catch {\n    return undefined;\n  }\n  return url;\n}\n\nconst LOOPBACK_HOSTNAMES = new Set([\"localhost\", \"127.0.0.1\", \"::1\", \"[::1]\"]);\n\n/** True for hostnames that always resolve to the local machine. */\nexport function isLoopbackHostname(host: string): boolean {\n  return LOOPBACK_HOSTNAMES.has(host);\n}\n\nfunction isLoopbackHttpUrl(url: URL): boolean {\n  return url.protocol === \"http:\" && isLoopbackHostname(url.hostname);\n}\n\n/** Read a response body as text, truncated to keep error messages bounded. */\nexport async function truncatedResponseText(response: Response, max = 500): Promise<string> {\n  const text = await response.text();\n  return text.slice(0, max);\n}\n\n/** Narrow an unknown value to a plain object, returning {} for arrays/primitives/null. */\nexport function asRecord(value: unknown): JsonObject {\n  return value && typeof value === \"object\" && !Array.isArray(value) ? (value as JsonObject) : {};\n}\n\n/** Extract a human-readable message from an unknown thrown value. */\nexport function errorMessage(error: unknown): string {\n  return error instanceof Error ? error.message : String(error);\n}\n\n/** Return the first finite number among the candidates, else undefined. */\nexport function firstNumber(...values: unknown[]): number | undefined {\n  for (const value of values) {\n    if (typeof value === \"number\" && Number.isFinite(value)) {\n      return value;\n    }\n  }\n  return undefined;\n}\n\n/** Generate a dash-free random identifier for synthesized response/message ids. */\nexport function randomId(): string {\n  return crypto.randomUUID().replaceAll(\"-\", \"\");\n}\n\n/** Drop keys whose value is undefined so they are omitted from JSON output. */\nexport function removeUndefined<T extends object>(value: T): T {\n  return Object.fromEntries(Object.entries(value).filter(([, v]) => v !== undefined)) as T;\n}\n\n/** Parse JSON, returning undefined instead of throwing on malformed input. */\nexport function safeJsonParse(text: string): unknown {\n  try {\n    return JSON.parse(text);\n  } catch {\n    return undefined;\n  }\n}\n\n/** Parse JSON into a plain object, returning undefined on malformed or non-object input. */\nexport function parseJsonObject(text: string): JsonObject | undefined {\n  try {\n    return asRecord(JSON.parse(text));\n  } catch {\n    return undefined;\n  }\n}\n\n/**\n * Extract de-duplicated model IDs from an OpenAI-style `/models` response (an\n * object carrying a `data` array, or a bare array of model objects).\n */\nexport function modelIdsFromResponse(body: unknown): string[] {\n  const record = asRecord(body);\n  const data = Array.isArray(record.data) ? record.data : Array.isArray(body) ? body : [];\n  const seen = new Set<string>();\n  const ids: string[] = [];\n  for (const model of data) {\n    const id = asRecord(model).id;\n    if (typeof id !== \"string\" || id.length === 0 || seen.has(id)) {\n      continue;\n    }\n    seen.add(id);\n    ids.push(id);\n  }\n  return ids;\n}\n\n/** Canonical set of accepted streaming-proxy modes, kept in sync with {@link StreamingProxyMode}. */\nexport const STREAMING_PROXY_MODES = [\n  \"auto\",\n  \"buffer\",\n  \"live\",\n] as const satisfies readonly StreamingProxyMode[];\n\n/** Validate a stream-mode string against the allowed {@link StreamingProxyMode} values. */\nexport function parseStreamingProxyMode(value: string): StreamingProxyMode {\n  if ((STREAMING_PROXY_MODES as readonly string[]).includes(value)) {\n    return value as StreamingProxyMode;\n  }\n  throw new Error(`Invalid stream mode: ${value}. Expected ${STREAMING_PROXY_MODES.join(\", \")}.`);\n}\n"],"mappings":";AAEA,SAAS,aAAa;AACtB,SAAS,aAAa,mBAAmB;;;ACAlC,SAAS,kBAAkB,OAAuB;AACvD,SAAO,MAAM,QAAQ,QAAQ,EAAE;AACjC;AAGO,SAAS,SAAS,OAA+C;AACtE,QAAM,UAAU,OAAO,KAAK;AAC5B,SAAO,UAAU,UAAU;AAC7B;AAYO,SAAS,sBACd,QACA,mBACA,mBAAmB,OACV;AACT,QAAM,MAAM,SAAS,MAAM;AAC3B,MAAI,CAAC,KAAK;AACR,WAAO;AAAA,EACT;AACA,MAAI,IAAI,YAAY,IAAI,YAAY,IAAI,UAAU,IAAI,MAAM;AAC1D,WAAO;AAAA,EACT;AACA,MAAI,IAAI,aAAa,MAAM,IAAI,aAAa,KAAK;AAC/C,WAAO;AAAA,EACT;AACA,MAAI,kBAAkB,GAAG,GAAG;AAC1B,WAAO;AAAA,EACT;AACA,MAAI,IAAI,aAAa,UAAU;AAC7B,WAAO;AAAA,EACT;AACA,QAAM,OAAO,IAAI,SAAS,YAAY;AACtC,SAAO,kBAAkB,SAAS,IAAI,KAAK;AAC7C;AAEA,SAAS,SAAS,QAAiC;AACjD,MAAI;AACJ,MAAI;AACF,UAAM,IAAI,IAAI,MAAM;AAAA,EACtB,QAAQ;AACN,WAAO;AAAA,EACT;AACA,SAAO;AACT;AAEA,IAAM,qBAAqB,oBAAI,IAAI,CAAC,aAAa,aAAa,OAAO,OAAO,CAAC;AAGtE,SAAS,mBAAmB,MAAuB;AACxD,SAAO,mBAAmB,IAAI,IAAI;AACpC;AAEA,SAAS,kBAAkB,KAAmB;AAC5C,SAAO,IAAI,aAAa,WAAW,mBAAmB,IAAI,QAAQ;AACpE;AAGA,eAAsB,sBAAsB,UAAoB,MAAM,KAAsB;AAC1F,QAAM,OAAO,MAAM,SAAS,KAAK;AACjC,SAAO,KAAK,MAAM,GAAG,GAAG;AAC1B;AAGO,SAAS,SAAS,OAA4B;AACnD,SAAO,SAAS,OAAO,UAAU,YAAY,CAAC,MAAM,QAAQ,KAAK,IAAK,QAAuB,CAAC;AAChG;AAGO,SAAS,aAAa,OAAwB;AACnD,SAAO,iBAAiB,QAAQ,MAAM,UAAU,OAAO,KAAK;AAC9D;AAGO,SAAS,eAAe,QAAuC;AACpE,aAAW,SAAS,QAAQ;AAC1B,QAAI,OAAO,UAAU,YAAY,OAAO,SAAS,KAAK,GAAG;AACvD,aAAO;AAAA,IACT;AAAA,EACF;AACA,SAAO;AACT;AAGO,SAAS,WAAmB;AACjC,SAAO,OAAO,WAAW,EAAE,WAAW,KAAK,EAAE;AAC/C;AAGO,SAAS,gBAAkC,OAAa;AAC7D,SAAO,OAAO,YAAY,OAAO,QAAQ,KAAK,EAAE,OAAO,CAAC,CAAC,EAAE,CAAC,MAAM,MAAM,MAAS,CAAC;AACpF;AAGO,SAAS,cAAc,MAAuB;AACnD,MAAI;AACF,WAAO,KAAK,MAAM,IAAI;AAAA,EACxB,QAAQ;AACN,WAAO;AAAA,EACT;AACF;AAGO,SAAS,gBAAgB,MAAsC;AACpE,MAAI;AACF,WAAO,SAAS,KAAK,MAAM,IAAI,CAAC;AAAA,EAClC,QAAQ;AACN,WAAO;AAAA,EACT;AACF;AAMO,SAAS,qBAAqB,MAAyB;AAC5D,QAAM,SAAS,SAAS,IAAI;AAC5B,QAAM,OAAO,MAAM,QAAQ,OAAO,IAAI,IAAI,OAAO,OAAO,MAAM,QAAQ,IAAI,IAAI,OAAO,CAAC;AACtF,QAAM,OAAO,oBAAI,IAAY;AAC7B,QAAM,MAAgB,CAAC;AACvB,aAAW,SAAS,MAAM;AACxB,UAAM,KAAK,SAAS,KAAK,EAAE;AAC3B,QAAI,OAAO,OAAO,YAAY,GAAG,WAAW,KAAK,KAAK,IAAI,EAAE,GAAG;AAC7D;AAAA,IACF;AACA,SAAK,IAAI,EAAE;AACX,QAAI,KAAK,EAAE;AAAA,EACb;AACA,SAAO;AACT;AAGO,IAAM,wBAAwB;AAAA,EACnC;AAAA,EACA;AAAA,EACA;AACF;AAGO,SAAS,wBAAwB,OAAmC;AACzE,MAAK,sBAA4C,SAAS,KAAK,GAAG;AAChE,WAAO;AAAA,EACT;AACA,QAAM,IAAI,MAAM,wBAAwB,KAAK,cAAc,sBAAsB,KAAK,IAAI,CAAC,GAAG;AAChG;;;AD/IA,IAAM,mBAAmB;AACzB,IAAM,oBAAoB;AAC1B,IAAM,gBAAgB;AACtB,IAAM,2BAA2B;AACjC,IAAM,iCAAiC;AACvC,IAAM,iBAAiB;AAAA,EACrB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAUO,SAAS,sBACd,MACA,MAAyB,QAAQ,KACf;AAClB,QAAM,UAAU,SAAS,IAAI,eAAe,KAAK;AAKjD,QAAM,SACJ,SAAS,IAAI,cAAc,KAAK,SAAS,IAAI,gBAAgB,KAAK,wBAAwB;AAC5F,QAAM,UAAU,SAAS,IAAI,gBAAgB,KAAK;AAClD,QAAM,QAAQ,SAAS,IAAI,YAAY,KAAK;AAC5C,QAAM,kBAAkB,SAAS,IAAI,6BAA6B,KAAK;AACvE,QAAM,sBAAsB,yBAAyB,IAAI,6BAA6B;AACtF,QAAM,sBAAsB;AAAA,IAC1B;AAAA,IACA,cAAc,KAAK,UAAU,OAAO,CAAC;AAAA,IACrC;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACA,MAAI,sBAAsB,GAAG;AAC3B,wBAAoB,KAAK,4BAA4B,mBAAmB,EAAE;AAAA,EAC5E;AACA,QAAM,iBAAiB,GAAG,oBAAoB,KAAK,IAAI,CAAC;AAExD,SAAO;AAAA,IACL,MAAM;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MAUJ;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,4BAA4B,cAAc;AAAA,MAC1C;AAAA,MACA;AAAA,MACA;AAAA,MACA,0BAA0B,KAAK,UAAU,eAAe,CAAC;AAAA,MACzD,GAAG;AAAA,IACL;AAAA,IACA;AAAA,IACA;AAAA,IACA,KAAK,gBAAgB;AAAA,MACnB,GAAG;AAAA,MACH,gBAAgB;AAAA,IAClB,CAAC;AAAA,IACD;AAAA,EACF;AACF;AAKA,SAAS,0BAAkC;AACzC,SAAO,UAAU,OAAO,WAAW,CAAC;AACtC;AAEA,SAAS,yBAAyB,UAAsC;AACtE,QAAM,MAAM,SAAS,QAAQ;AAC7B,MAAI,QAAQ,QAAW;AACrB,WAAO;AAAA,EACT;AACA,QAAM,QAAQ,OAAO,GAAG;AACxB,MAAI,CAAC,OAAO,UAAU,KAAK,KAAK,QAAQ,GAAG;AACzC,UAAM,IAAI,MAAM,+DAA+D;AAAA,EACjF;AACA,SAAO;AACT;AAEA,SAAS,gBAAgB,KAA2C;AAClE,QAAM,OAAO,EAAE,GAAG,IAAI;AACtB,aAAW,OAAO,gBAAgB;AAChC,WAAO,KAAK,GAAG;AAAA,EACjB;AACA,SAAO;AACT;AAEA,eAAsB,KAAK,OAAO,IAAI,KAAK,MAAM,CAAC,GAAG,MAAM,QAAQ,KAAoB;AACrF,MAAI,KAAK,WAAW,MAAM,KAAK,CAAC,MAAM,YAAY,KAAK,CAAC,MAAM,OAAO;AACnE,YAAQ,IAAI,SAAS,CAAC;AACtB;AAAA,EACF;AAEA,QAAM,aAAa,sBAAsB,MAAM,GAAG;AAClD,MAAI,IAAI,gCAAgC,KAAK;AAC3C,UAAM,kBAAkB,UAAU;AAAA,EACpC;AACA,QAAM,QAAQ,MAAM,WAAW,SAAS,WAAW,MAAM;AAAA,IACvD,KAAK,WAAW;AAAA,IAChB,OAAO,QAAQ,aAAa;AAAA,IAC5B,OAAO;AAAA,EACT,CAAC;AAED,QAAM,WAAW,MAAM,IAAI,QAAgB,CAAC,SAAS,WAAW;AAC9D,UAAM,KAAK,SAAS,MAAM;AAC1B,UAAM,KAAK,QAAQ,CAAC,MAAM,WAAW;AACnC,UAAI,OAAO,SAAS,UAAU;AAC5B,gBAAQ,IAAI;AACZ;AAAA,MACF;AACA,cAAQ,SAAS,MAAM,aAAa,MAAM,IAAI,CAAC;AAAA,IACjD,CAAC;AAAA,EACH,CAAC;AAED,UAAQ,WAAW;AACrB;AAEA,eAAsB,kBACpB,YACA,UAAqB,OACN;AACf,QAAM,YAAY,GAAG,kBAAkB,WAAW,OAAO,CAAC;AAC1D,QAAM,SAAS,WAAW,IAAI;AAC9B,MAAI,WAAW,QAAW;AACxB,UAAM,IAAI;AAAA,MACR;AAAA,IACF;AAAA,EACF;AACA,MAAI;AACJ,MAAI;AACF,eAAW,MAAM,QAAQ,WAAW;AAAA,MAClC,SAAS;AAAA,QACP,QAAQ;AAAA,QACR,eAAe,UAAU,MAAM;AAAA,MACjC;AAAA,MACA,QAAQ;AAAA,IACV,CAAC;AAAA,EACH,SAAS,OAAO;AACd,UAAM,IAAI;AAAA,MACR,+BAA+B,SAAS,oFAAoF,aAAa,KAAK,CAAC;AAAA,IACjJ;AAAA,EACF;AAEA,MAAI,CAAC,SAAS,IAAI;AAChB,UAAM,IAAI;AAAA,MACR,0BAA0B,KAAK,UAAU,WAAW,KAAK,CAAC,YAAY,SAAS,aAAa,SAAS,MAAM,KAAK,MAAM,sBAAsB,QAAQ,CAAC;AAAA,IACvJ;AAAA,EACF;AAEA,QAAM,SAAS,qBAAqB,MAAM,SAAS,KAAK,EAAE,MAAM,MAAM,MAAS,CAAC;AAChF,MAAI,OAAO,SAAS,KAAK,CAAC,OAAO,SAAS,WAAW,KAAK,GAAG;AAC3D,UAAM,IAAI;AAAA,MACR,0DAA0D,KAAK,UAAU,WAAW,KAAK,CAAC,OAAO,SAAS,uBAAuB,OAAO,KAAK,IAAI,CAAC;AAAA,IACpJ;AAAA,EACF;AACF;AAEA,SAAS,WAAmB;AAC1B,SAAO;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8DAQqD,gBAAgB;AAAA;AAAA;AAAA;AAAA;AAAA,2DAKnB,iBAAiB;AAAA,sDACtB,aAAa;AAAA;AAAA,0DAET,wBAAwB;AAAA;AAAA;AAAA,yBAGzD,8BAA8B;AAAA;AAAA;AAAA;AAAA,qKAI8G,aAAa,SAAS,wBAAwB;AACnN;AAEA,SAAS,aAAa,QAAgC;AACpD,SAAO,YAAY,QAAQ,MAAM,KAAK;AACxC;AAEA,IAAI,YAAY,MAAM;AACpB,OAAK,EAAE,MAAM,CAAC,UAAmB;AAC/B,YAAQ,MAAM,aAAa,KAAK,CAAC;AACjC,YAAQ,KAAK,CAAC;AAAA,EAChB,CAAC;AACH;","names":[]}

package/dist/cli.js

@@ -15,7 +15,7 @@ import {
   safeJsonParse,
   trimTrailingSlash,
   truncatedResponseText
-} from "./chunk-CYR6I4C3.js";
+} from "./chunk-4ZG5QEYJ.js";
 
 // src/cli.ts
 import { spawn } from "child_process";
@@ -176,6 +176,14 @@ var COPILOT_USAGE_API_VERSION = "2025-04-01";
 var EDITOR_PLUGIN_VERSION = "hoopilot/0.1.0";
 var EDITOR_VERSION = "Hoopilot/0.1.0";
 var HOOPILOT_USER_AGENT = "hoopilot/0.1.0";
+var DEFAULT_UPSTREAM_TIMEOUT_MS = 12e4;
+var DEFAULT_UPSTREAM_STREAM_IDLE_TIMEOUT_MS = 12e4;
+var CopilotUpstreamTimeoutError = class extends Error {
+  constructor(message) {
+    super(message);
+    this.name = "CopilotUpstreamTimeoutError";
+  }
+};
 function applyCopilotHeaders(headers, token) {
   headers.set("accept", headers.get("accept") ?? "application/json");
   headers.set("authorization", `Bearer ${token}`);
@@ -228,6 +236,8 @@ var CopilotClient = class {
   #allowUnsafeUpstream;
   #fetch;
   #githubApiBaseUrl;
+  #upstreamStreamIdleTimeoutMs;
+  #upstreamTimeoutMs;
   constructor(options = {}) {
     this.#auth = new CopilotAuth(options);
     this.#allowUnsafeUpstream = envValue(options.env?.HOOPILOT_ALLOW_UNSAFE_UPSTREAM) === "1";
@@ -235,6 +245,18 @@ var CopilotClient = class {
     this.#githubApiBaseUrl = trimTrailingSlash(
       options.githubApiBaseUrl ?? envValue(options.env?.HOOPILOT_GITHUB_API_BASE_URL) ?? DEFAULT_GITHUB_API_BASE_URL
     );
+    this.#upstreamTimeoutMs = parseTimeoutMs(
+      options.upstreamTimeoutMs,
+      options.env?.HOOPILOT_UPSTREAM_TIMEOUT_MS,
+      DEFAULT_UPSTREAM_TIMEOUT_MS,
+      "HOOPILOT_UPSTREAM_TIMEOUT_MS"
+    );
+    this.#upstreamStreamIdleTimeoutMs = parseTimeoutMs(
+      options.upstreamStreamIdleTimeoutMs,
+      options.env?.HOOPILOT_UPSTREAM_STREAM_IDLE_TIMEOUT_MS,
+      DEFAULT_UPSTREAM_STREAM_IDLE_TIMEOUT_MS,
+      "HOOPILOT_UPSTREAM_STREAM_IDLE_TIMEOUT_MS"
+    );
   }
   /**
    * Fetch the Copilot account's quota / premium-request usage from the GitHub
@@ -253,7 +275,7 @@ var CopilotClient = class {
     }
     const access = await this.#auth.getAccess();
     const headers = applyGithubApiHeaders(new Headers(), access.token);
-    return this.#fetch(`${this.#githubApiBaseUrl}/copilot_internal/user`, {
+    return this.#fetchWithTimeout(`${this.#githubApiBaseUrl}/copilot_internal/user`, {
       headers,
       method: "GET",
       signal
@@ -300,12 +322,139 @@ var CopilotClient = class {
       );
     }
     const headers = applyCopilotHeaders(new Headers(init.headers), access.token);
-    return this.#fetch(`${access.apiBaseUrl}${path}`, {
+    return this.#fetchWithTimeout(`${access.apiBaseUrl}${path}`, {
       ...init,
       headers
     });
   }
+  async #fetchWithTimeout(input, init) {
+    const timeout = abortSignalWithTimeout(init.signal ?? void 0, this.#upstreamTimeoutMs);
+    try {
+      const response = await this.#fetch(input, {
+        ...init,
+        signal: timeout.signal
+      });
+      return responseWithStreamIdleTimeout(response, this.#upstreamStreamIdleTimeoutMs, input);
+    } catch (error) {
+      if (timeout.timedOut()) {
+        throw new CopilotUpstreamTimeoutError(
+          `Copilot upstream request timed out after ${this.#upstreamTimeoutMs} ms before response headers arrived.`
+        );
+      }
+      throw error;
+    } finally {
+      timeout.cleanup();
+    }
+  }
 };
+function parseTimeoutMs(optionValue2, envRaw, fallback, name) {
+  const raw = optionValue2 ?? envValue(envRaw);
+  if (raw === void 0) {
+    return fallback;
+  }
+  const value = typeof raw === "number" ? raw : Number(raw);
+  if (!Number.isInteger(value) || value < 0) {
+    throw new Error(`${name} must be a non-negative integer number of milliseconds.`);
+  }
+  return value;
+}
+function abortSignalWithTimeout(parent, timeoutMs) {
+  if (timeoutMs === 0) {
+    return { cleanup: () => {
+    }, signal: parent, timedOut: () => false };
+  }
+  const controller = new AbortController();
+  let timedOut = false;
+  const timer = setTimeout(() => {
+    if (controller.signal.aborted) {
+      return;
+    }
+    timedOut = true;
+    controller.abort(
+      new CopilotUpstreamTimeoutError(`Copilot upstream request timed out after ${timeoutMs} ms.`)
+    );
+  }, timeoutMs);
+  const onAbort = () => controller.abort(parent?.reason);
+  if (parent?.aborted) {
+    controller.abort(parent.reason);
+  } else {
+    parent?.addEventListener("abort", onAbort, { once: true });
+  }
+  return {
+    cleanup: () => {
+      clearTimeout(timer);
+      parent?.removeEventListener("abort", onAbort);
+    },
+    signal: controller.signal,
+    timedOut: () => timedOut
+  };
+}
+function responseWithStreamIdleTimeout(response, idleTimeoutMs, input) {
+  if (!response.body || idleTimeoutMs === 0) {
+    return response;
+  }
+  return new Response(streamWithIdleTimeout(response.body, idleTimeoutMs, input), {
+    headers: response.headers,
+    status: response.status,
+    statusText: response.statusText
+  });
+}
+function streamWithIdleTimeout(body, idleTimeoutMs, input) {
+  const reader = body.getReader();
+  let released = false;
+  const release = () => {
+    if (!released) {
+      released = true;
+      reader.releaseLock();
+    }
+  };
+  return new ReadableStream({
+    async pull(controller) {
+      let timer;
+      const read = reader.read();
+      read.catch(() => {
+      });
+      try {
+        const result = await Promise.race([
+          read,
+          new Promise((_, reject) => {
+            timer = setTimeout(() => {
+              reject(
+                new CopilotUpstreamTimeoutError(
+                  `Copilot upstream stream was idle for ${idleTimeoutMs} ms while reading ${input}.`
+                )
+              );
+            }, idleTimeoutMs);
+          })
+        ]);
+        if (timer) {
+          clearTimeout(timer);
+        }
+        if (result.done) {
+          controller.close();
+          release();
+          return;
+        }
+        controller.enqueue(result.value);
+      } catch (error) {
+        if (timer) {
+          clearTimeout(timer);
+        }
+        await reader.cancel(error).catch(() => {
+        });
+        controller.error(error);
+        release();
+      }
+    },
+    async cancel(reason) {
+      try {
+        await reader.cancel(reason);
+      } finally {
+        release();
+      }
+    }
+  });
+}
 function normalizeCopilotUsage(body) {
   const record = asRecord(body);
   const quotas = {};
@@ -3139,17 +3288,15 @@ function observeResponseUsage(response, fallbackModel, onUsage, signal, onOutcom
   if (!body) {
     return response;
   }
-  const [clientBranch, observerBranch] = body.tee();
   const isSse = response.headers.get("content-type")?.includes("text/event-stream") ?? false;
-  void consumeUsage(observerBranch, isSse, fallbackModel, onUsage, signal, onOutcome).catch(
-    () => {
+  return new Response(
+    streamWithUsageObservation(body, isSse, fallbackModel, onUsage, signal, onOutcome),
+    {
+      headers: response.headers,
+      status: response.status,
+      statusText: response.statusText
     }
   );
-  return new Response(clientBranch, {
-    headers: response.headers,
-    status: response.status,
-    statusText: response.statusText
-  });
 }
 function recordResponseTextUsage(text, isSse, fallbackModel, onUsage, onOutcome) {
   const accumulator = createUsageAccumulator(fallbackModel, onUsage, onOutcome);
@@ -3165,13 +3312,16 @@ function recordResponseTextUsage(text, isSse, fallbackModel, onUsage, onOutcome)
   }
   accumulator.finish();
 }
-async function consumeUsage(stream, isSse, fallbackModel, onUsage, signal, onOutcome) {
+function streamWithUsageObservation(stream, isSse, fallbackModel, onUsage, signal, onOutcome) {
   const reader = stream.getReader();
+  let aborted = signal?.aborted ?? false;
+  let released = false;
   const onAbort = () => {
+    aborted = true;
     reader.cancel().catch(() => {
     });
   };
-  if (signal?.aborted) {
+  if (aborted) {
     reader.cancel().catch(() => {
     });
   } else {
@@ -3179,7 +3329,7 @@ async function consumeUsage(stream, isSse, fallbackModel, onUsage, signal, onOut
   }
   const decoder = new TextDecoder();
   const guardedOutcome = onOutcome ? (extracted) => {
-    if (!signal?.aborted) {
+    if (!aborted) {
       onOutcome(extracted);
     }
   } : void 0;
@@ -3187,33 +3337,40 @@ async function consumeUsage(stream, isSse, fallbackModel, onUsage, signal, onOut
   let buffer = "";
   let bufferedBytes = 0;
   let overflowed = false;
-  try {
-    while (true) {
-      const result = await reader.read();
-      if (result.done) {
-        break;
+  const release = () => {
+    if (released) {
+      return;
+    }
+    released = true;
+    signal?.removeEventListener("abort", onAbort);
+    reader.releaseLock();
+  };
+  const observeChunk = (chunkBytes) => {
+    const chunk = decoder.decode(chunkBytes, { stream: true });
+    if (isSse) {
+      buffer += chunk;
+      const lines = buffer.split(/\r?\n/);
+      buffer = lines.pop() ?? "";
+      for (const line of lines) {
+        considerSseLine(line, accumulator.consider);
       }
-      const chunk = decoder.decode(result.value, { stream: true });
-      if (isSse) {
-        buffer += chunk;
-        const lines = buffer.split(/\r?\n/);
-        buffer = lines.pop() ?? "";
-        for (const line of lines) {
-          considerSseLine(line, accumulator.consider);
-        }
-        if (buffer.length > USAGE_BUFFER_LIMIT_BYTES) {
-          buffer = "";
-        }
-      } else if (!overflowed) {
-        bufferedBytes += result.value.byteLength;
-        if (bufferedBytes > USAGE_BUFFER_LIMIT_BYTES) {
-          overflowed = true;
-          buffer = "";
-        } else {
-          buffer += chunk;
-        }
+      if (buffer.length > USAGE_BUFFER_LIMIT_BYTES) {
+        buffer = "";
       }
+      return;
     }
+    if (overflowed) {
+      return;
+    }
+    bufferedBytes += chunkBytes.byteLength;
+    if (bufferedBytes > USAGE_BUFFER_LIMIT_BYTES) {
+      overflowed = true;
+      buffer = "";
+      return;
+    }
+    buffer += chunk;
+  };
+  const finishObservation = () => {
     const finalBuffer = buffer + decoder.decode();
     if (isSse) {
       if (finalBuffer) {
@@ -3225,11 +3382,41 @@ async function consumeUsage(stream, isSse, fallbackModel, onUsage, signal, onOut
         accumulator.consider(parsed);
       }
     }
-  } finally {
-    signal?.removeEventListener("abort", onAbort);
-    reader.releaseLock();
-  }
-  accumulator.finish();
+    if (!aborted) {
+      safeFinishAccumulator(accumulator);
+    }
+  };
+  return new ReadableStream({
+    async pull(controller) {
+      const result = await reader.read().catch((error) => {
+        release();
+        controller.error(error);
+        return void 0;
+      });
+      if (!result) {
+        return;
+      }
+      if (result.done) {
+        finishObservation();
+        controller.close();
+        release();
+        return;
+      }
+      try {
+        observeChunk(result.value);
+      } catch {
+      }
+      controller.enqueue(result.value);
+    },
+    async cancel(reason) {
+      aborted = true;
+      try {
+        await reader.cancel(reason);
+      } finally {
+        release();
+      }
+    }
+  });
 }
 function createUsageAccumulator(fallbackModel, onUsage, onOutcome) {
   let model = fallbackModel;
@@ -3254,6 +3441,12 @@ function createUsageAccumulator(fallbackModel, onUsage, onOutcome) {
     }
   };
 }
+function safeFinishAccumulator(accumulator) {
+  try {
+    accumulator.finish();
+  } catch {
+  }
+}
 function considerSseLine(line, consider) {
   const trimmed = line.trim();
   if (!trimmed.startsWith("data:")) {
@@ -3372,7 +3565,18 @@ async function getVersion() {
 var DEFAULT_HOST = "127.0.0.1";
 var DEFAULT_PORT = 4141;
 var FORBIDDEN_BROWSER_ORIGIN_MESSAGE = "Cross-origin browser requests are blocked unless the Origin is loopback or listed in HOOPILOT_ALLOWED_ORIGINS.";
-var WELL_KNOWN_DEMO_API_KEYS = /* @__PURE__ */ new Set(["local-key"]);
+var MIN_NON_LOOPBACK_API_KEY_LENGTH = 24;
+var WELL_KNOWN_DEMO_API_KEYS = /* @__PURE__ */ new Set([
+  "changeme",
+  "demo",
+  "example",
+  "hoopilot",
+  "local-key",
+  "password",
+  "password123",
+  "secret",
+  "test"
+]);
 var INVALID_JSON_MESSAGE = "Request body must be valid JSON.";
 var JSON_OBJECT_MESSAGE = "Request body must be a JSON object.";
 var MAX_REQUEST_BODY_BYTES = 16 * 1024 * 1024;
@@ -3545,6 +3749,13 @@ function buildApp(deps) {
       );
       return jsonError(413, "request_too_large", message);
     }
+    if (error instanceof CopilotUpstreamTimeoutError) {
+      logger.warn(
+        { err: errorDetails(error), event: "copilot.request.timeout" },
+        "copilot upstream request timed out"
+      );
+      return jsonError(504, "copilot_timeout", message);
+    }
     logger.error({ err: errorDetails(error), event: "http.request.failed" }, "request failed");
     return jsonError(500, "internal_error", message);
   }).get("/", () => jsonResponse({ name: "hoopilot", object: "health", status: "ok" })).get("/healthz", () => jsonResponse({ name: "hoopilot", object: "health", status: "ok" })).get("/metrics", () => metricsResponse(metrics)).get("/v1/usage", ({ request }) => handleUsage(metrics, readUsage, request.signal)).get(
@@ -3643,10 +3854,9 @@ function startHoopilotServer(options = {}) {
         "Refusing to listen on a non-loopback host without HOOPILOT_API_KEY. Set an API key or pass --allow-unauthenticated."
       );
     }
-    if (apiKey && isWellKnownDemoApiKey(apiKey)) {
-      throw new Error(
-        "Refusing to listen on a non-loopback host with a well-known demo HOOPILOT_API_KEY. Set a strong, unique API key."
-      );
+    const rejection = apiKey ? apiKeyRejectionReason(apiKey) : void 0;
+    if (rejection) {
+      throw new Error(`Refusing to listen on a non-loopback host: ${rejection}`);
     }
   }
   const server = Bun.serve({
@@ -3944,12 +4154,16 @@ async function readRequestText(request) {
   const reader = body.getReader();
   const decoder = new TextDecoder();
   let bytes = 0;
-  let text = "";
+  const chunks = [];
   try {
     while (true) {
       const { done, value } = await reader.read();
       if (done) {
-        return `${text}${decoder.decode()}`;
+        const tail = decoder.decode();
+        if (tail) {
+          chunks.push(tail);
+        }
+        return chunks.join("");
       }
       bytes += value.byteLength;
       if (bytes > MAX_REQUEST_BODY_BYTES) {
@@ -3957,7 +4171,7 @@ async function readRequestText(request) {
         });
         throw new RequestBodyTooLargeError();
       }
-      text += decoder.decode(value, { stream: true });
+      chunks.push(decoder.decode(value, { stream: true }));
     }
   } finally {
     reader.releaseLock();
@@ -4054,8 +4268,18 @@ function resolveCorsAllowOrigin(origin, allowedOrigins) {
   }
   return isAllowedOrigin(origin, allowedOrigins) ? origin : void 0;
 }
-function isWellKnownDemoApiKey(apiKey) {
-  return WELL_KNOWN_DEMO_API_KEYS.has(apiKey.trim().toLowerCase());
+function apiKeyRejectionReason(apiKey) {
+  const normalized = apiKey.trim();
+  if (WELL_KNOWN_DEMO_API_KEYS.has(normalized.toLowerCase())) {
+    return "HOOPILOT_API_KEY is a well-known demo value. Set a strong, unique API key.";
+  }
+  if (normalized.length < MIN_NON_LOOPBACK_API_KEY_LENGTH) {
+    return `HOOPILOT_API_KEY must be at least ${MIN_NON_LOOPBACK_API_KEY_LENGTH} characters when listening on a non-loopback host.`;
+  }
+  if (/^(.)\1+$/.test(normalized)) {
+    return "HOOPILOT_API_KEY must not be a repeated single character. Set a strong, unique API key.";
+  }
+  return void 0;
 }
 function isUpstreamAuthStatus(status) {
   return status === 401 || status === 403;
@@ -4435,8 +4659,14 @@ function versionFromTag(tag) {
   return tag.trim().replace(/^v/, "");
 }
 function assetSuffixFor(platform, arch, isMusl) {
-  const os = platform === "win32" ? "windows" : platform === "darwin" ? "darwin" : "linux";
-  const cpu = arch === "arm64" || arch === "aarch64" ? "arm64" : "x64";
+  const os = platform === "linux" ? "linux" : platform === "win32" ? "windows" : platform === "darwin" ? "darwin" : void 0;
+  if (!os) {
+    throw new Error(`Unsupported platform for standalone updates: ${platform}.`);
+  }
+  const cpu = arch === "x64" || arch === "amd64" ? "x64" : arch === "arm64" || arch === "aarch64" ? "arm64" : void 0;
+  if (!cpu) {
+    throw new Error(`Unsupported architecture for standalone updates: ${arch}.`);
+  }
   const libc = os === "linux" && isMusl ? "-musl" : "";
   return `${os}-${cpu}${libc}`;
 }
@@ -5325,6 +5555,7 @@ Options:
   -p, --port <port>                 Port to listen on. Default: 4141
       --host <host>                 Host to listen on. Default: 127.0.0.1
       --api-key <key>               Require clients to send Authorization: Bearer <key> or x-api-key: <key>
+                                    Non-loopback binds require at least 24 characters.
       --api-key-file <path>         Read the local API key from a file instead of argv
       --auth-file <path>            OAuth credential store path
       --copilot-api-base-url <url>  Copilot API base URL override