@zhiman_innies/innies-codex 0.122.50 → 0.122.54

package/bin/innies-config.js

@@ -21,7 +21,6 @@ const DASHSCOPE_MODEL = "qwen3.6-27b";
 const PRIVATE_27B_MODEL = "qwen36_27b";
 const PRIVATE_27B_PROVIDER = "zhiman_27b";
 const QWEN_MODELS = new Set([DEFAULT_MODEL, DASHSCOPE_MODEL, PRIVATE_27B_MODEL]);
-const LEGACY_MODELS = new Set(["qwen-plus", "qwen-plus-latest"]);
 const DEFAULT_HOME_DIR = ".inniescoder";
 const DEFAULT_CATALOG_FILENAME = "catalog.json";
 const DEFAULT_CONFIG_FILENAME = "config.toml";
@@ -237,12 +236,12 @@ function ensureInniesCatalog(catalogPath) {
     return;
   }
 
-  if (shouldReplaceLegacyCatalog(existingCatalog)) {
+  if (catalogShapeIsStale(existingCatalog)) {
     writeJsonFile(catalogPath, defaultCatalog);
   }
 }
 
-function shouldReplaceLegacyCatalog(catalog) {
+function catalogShapeIsStale(catalog) {
   if (!catalog || !Array.isArray(catalog.models)) {
     return false;
   }
@@ -278,7 +277,7 @@ function shouldReplaceLegacyCatalog(catalog) {
     return true;
   }
 
-  return slugs.some((slug) => LEGACY_MODELS.has(slug));
+  return false;
 }
 
 function ensureInniesConfig(configPath, catalogPath, state) {
@@ -300,15 +299,33 @@ function defaultInniesConfig(catalogPath, managedDefault) {
   const lines = managedDefault
     ? managedDefaultLines(catalogPath, managedDefault)
     : [`model_catalog_json = ${JSON.stringify(catalogPath)}`];
+  // The documentation comment block (model_reasoning_effort /
+  // model_verbosity / etc.) is emitted AFTER the active root settings
+  // — this matches the byte ordering that `normalizeInniesConfig`
+  // produces on re-init (which prepends active managed lines and
+  // appends the rest as unmanaged content, including any preserved
+  // comments). With this ordering, init 1's output is bytewise equal
+  // to init 2+'s normalized output, preserving idempotency. The
+  // comments are preserved across re-inits because
+  // `stripManagedRootSettings` only strips active `key = ...` lines,
+  // not commented ones.
+  const rootDocumentation = [
+    `# Managed defaults — override by uncommenting and editing.`,
+    `# innies-config.js will keep these in sync on every \`innies\` run,`,
+    `# but will NEVER overwrite a user-supplied value.`,
+    `# model_reasoning_effort = "low"   # thinking toggle: low | medium | high | xhigh (or omit for model default)`,
+    `# model_verbosity = "low"           # optional: GPT-5 only (low | medium | high)`,
+  ];
   return [
     ...lines,
     "",
+    ...rootDocumentation,
+    "",
     defaultZhiman35bProviderBlock(),
     "",
     defaultZhiman27bProviderBlock(),
     "",
     defaultDashscopeProviderBlock(),
-    "",
   ].join("\n");
 }
 
@@ -337,19 +354,26 @@ function normalizeInniesConfig(contents, catalogPath, state) {
 
   updated = normalizeManagedProviderBlocks(stripReservedProviderBlocks(updated));
   if (!updated.includes(ZHIMAN_35B_PROVIDER_HEADER)) {
-    updated = `${updated.trimEnd()}\n\n${defaultZhiman35bProviderBlock()}\n`;
+    updated = `${updated.trimEnd()}\n\n${defaultZhiman35bProviderBlock()}`;
   }
   if (!updated.includes(ZHIMAN_27B_PROVIDER_HEADER)) {
-    updated = `${updated.trimEnd()}\n\n${defaultZhiman27bProviderBlock()}\n`;
+    updated = `${updated.trimEnd()}\n\n${defaultZhiman27bProviderBlock()}`;
   }
   if (!updated.includes(DASHSCOPE_PROVIDER_HEADER)) {
-    updated = `${updated.trimEnd()}\n\n${defaultDashscopeProviderBlock()}\n`;
+    updated = `${updated.trimEnd()}\n\n${defaultDashscopeProviderBlock()}`;
   }
 
   return updated;
 }
 
 function managedDefaultLines(catalogPath, managedDefault) {
+  // IMPORTANT: only ACTIVE settings belong here. Comment lines
+  // (`# model_reasoning_effort = "..."`) belong in
+  // `defaultInniesConfig` (one-time generation) — they would
+  // otherwise be PREPENDED on every re-init and accumulate forever,
+  // breaking bytewise idempotency. `stripManagedRootSettings` only
+  // strips active `key = ...` lines (not commented ones), so anything
+  // we emit here MUST be active (no leading `#`).
   return [
     `model_provider = ${JSON.stringify(managedDefault.provider)}`,
     `model = ${JSON.stringify(managedDefault.model)}`,
@@ -358,7 +382,9 @@ function managedDefaultLines(catalogPath, managedDefault) {
 }
 
 function catalogOnlyManagedLines(catalogPath) {
-  return [`model_catalog_json = ${JSON.stringify(catalogPath)}`];
+  return [
+    `model_catalog_json = ${JSON.stringify(catalogPath)}`,
+  ];
 }
 
 function preservedUserManagedLines(contents, catalogPath) {
@@ -424,23 +450,36 @@ function determineModelSelectionState(contents, previousState) {
     return previousState;
   }
 
+  // Goal 4: model switching must NOT cause local-config corruption.
+  // When the user has actively chosen one of the supported NON-DEFAULT
+  // models (DASHSCOPE_MODEL=qwen3.6-27b, PRIVATE_27B_MODEL=qwen36_27b),
+  // we lock the state to USER_SELECTED so subsequent `innies` runs
+  // never strip-and-rewrite their model_provider/model with the managed
+  // default qwen35_35b. The previous behaviour returned
+  // `previousState` here, which silently reset the user's selection
+  // when innies-state.json was missing or corrupted (e.g. after a
+  // version upgrade that lost the state file) — see manual e2e test
+  // `state-loss reset` in this commit's changelog.
   if (
     currentModel === DASHSCOPE_MODEL &&
     (currentProvider == null || currentProvider === "dashscope")
   ) {
-    return previousState;
+    return {
+      model_selection_state: MODEL_SELECTION_STATES.USER_SELECTED,
+    };
   }
 
   if (
     currentModel === PRIVATE_27B_MODEL &&
     (currentProvider == null || currentProvider === PRIVATE_27B_PROVIDER)
   ) {
-    return previousState;
+    return {
+      model_selection_state: MODEL_SELECTION_STATES.USER_SELECTED,
+    };
   }
 
   if (
     currentModel != null &&
-    !LEGACY_MODELS.has(currentModel) &&
     (currentModel !== DEFAULT_MODEL || currentProvider !== DEFAULT_PROVIDER) &&
     (currentModel !== DASHSCOPE_MODEL || currentProvider !== "dashscope") &&
     (currentModel !== PRIVATE_27B_MODEL || currentProvider !== PRIVATE_27B_PROVIDER)
@@ -492,11 +531,16 @@ function normalizeManagedProviderBlocks(contents) {
   // NOTE: we deliberately do NOT auto-inject `base_url` or `env_key`
   // into user blocks. The user must configure these themselves (either
   // in the TOML block or via the corresponding env var: ZHIMAN_API_KEY
-  // / DASHSCOPE_API_KEY / ZHIMAN_35B_API_KEY). The Rust builtin
-  // providers fall back to the env var when `env_key` is absent from
-  // the TOML, so the only thing that is strictly required from the
-  // user is `base_url` (or the ZHIMAN_BASE_URL / DASHSCOPE_BASE_URL env
-  // var). We only normalize `wire_api` to a known-good value.
+  // / DASHSCOPE_API_KEY). The Rust builtin providers fall back to the
+  // env var when `env_key` is absent from the TOML, so the only thing
+  // that is strictly required from the user is `base_url` (or the
+  // ZHIMAN_BASE_URL / DASHSCOPE_BASE_URL env var). We only normalize
+  // `wire_api` to a known-good value. (Historically the comment also
+  // mentioned `ZHIMAN_35B_API_KEY`, but the Rust factory at
+  // `codex-rs/model-provider-info/src/lib.rs` hardcodes a single
+  // `ZHIMAN_API_KEY` env-var name for ALL `zhiman_*` provider blocks —
+  // there is no per-model differentiation. Users who need separate
+  // env vars must override `env_key` in the TOML block directly.)
   let updated = normalizeProviderBlock(contents, {
     providerHeader: ZHIMAN_35B_PROVIDER_HEADER,
   });
@@ -610,22 +654,53 @@ function normalizeProviderBlock(contents, provider) {
 }
 
 function defaultZhiman35bProviderBlock() {
-  // The freshly-generated provider block intentionally does NOT include
-  // `base_url` or `env_key`. The user must configure both:
+  // The freshly-generated provider block intentionally does NOT prefill
+  // `base_url` or `env_key` (commented placeholders only). The user must
+  // configure both:
   //
   //   base_url = "http://your-private-deployment/v1"
-  //   env_key  = "ZHIMAN_35B_API_KEY"   # or any other env var name
+  //   env_key  = "ZHIMAN_API_KEY"   # or any other env var name
   //
-  // (or set the ZHIMAN_BASE_URL and ZHIMAN_35B_API_KEY environment
+  // (or set the ZHIMAN_BASE_URL and ZHIMAN_API_KEY environment
   // variables). Generating a default `base_url` here would make the
   // binary call a real network endpoint on first run, which is
   // explicitly disallowed by the innies-codex install contract.
+  //
+  // The block also exposes ALL recognized ModelProviderInfo fields as
+  // commented placeholders so the operator can see what knobs are
+  // available without reading the upstream schema:
+  //   - http_headers / env_http_headers: per-request headers (literal
+  //     or env-var indirection)
+  //   - query_params: extra URL query params
+  //   - request_max_retries / stream_max_retries: retry budgets
+  //   - stream_idle_timeout_ms: per-stream idle cap (defends against
+  //     hung first-token — see client.rs::ZHIMAN_STREAM_REQUEST_TIMEOUT)
+  //   - request_body_extras: free-form JSON merged into every
+  //     chat-completions request body (qwen uses this for the nested
+  //     `chat_template_kwargs` thinking toggle)
+  //
+  // NOTE: the env-var name is `ZHIMAN_API_KEY` (NOT `ZHIMAN_35B_API_KEY`)
+  // because the Rust builtin factory at
+  // `codex-rs/model-provider-info/src/lib.rs` hardcodes that single
+  // env-var name for ALL `zhiman_*` provider blocks — there is no
+  // per-model env-var differentiation. If you need different keys for
+  // the 27B vs 35B deployments, override `env_key` in the TOML block
+  // directly (e.g. `env_key = "ZHIMAN_35B_API_KEY"`) instead of relying
+  // on the factory default.
   return [
     ZHIMAN_35B_PROVIDER_HEADER,
     'name = "zhiman_35b"',
     `# base_url = "http://your-private-deployment/v1"   # FILL IN: private vLLM / OpenAI-compatible endpoint`,
-    `# env_key  = "ZHIMAN_35B_API_KEY"                   # FILL IN: name of the env var holding your API key`,
+    `# env_key  = "ZHIMAN_API_KEY"                      # FILL IN: name of the env var holding your API key`,
     `wire_api = "${DEFAULT_PROVIDER_WIRE_API}"`,
+    `# http_headers = { "X-Custom" = "value" }          # optional: literal per-request headers`,
+    `# env_http_headers = { "X-Tenant" = "TENANT_ID_ENV" }  # optional: headers from env-var indirection`,
+    `# query_params = { "api-version" = "2024-10" }     # optional: extra URL query params`,
+    `# request_max_retries = 4                          # optional: HTTP retry budget`,
+    `# stream_max_retries = 5                           # optional: SSE stream reconnect retry budget`,
+    `# stream_idle_timeout_ms = 300000                   # optional: idle cap (default 300_000 = 5 min)`,
+    `# request_body_extras = { chat_template_kwargs = { enable_thinking = false } }  # optional: qwen thinking toggle (private vLLM)`,
+    `# experimental_bearer_token = ""                   # discouraged: prefer env_key`,
   ].join("\n");
 }
 
@@ -634,14 +709,25 @@ function defaultZhiman27bProviderBlock() {
   // deployment. Same install-contract rule: do NOT prefill `base_url` or
   // `env_key` — the user must configure both, otherwise the binary
   // would call a real network endpoint on first run. Distinct provider
-  // name (`zhiman_27b` vs `zhiman_35b`) and env var so the two private
-  // deployments can be configured independently.
+  // name (`zhiman_27b` vs `zhiman_35b`) so the two private deployments
+  // can be configured independently. NOTE: the env-var name is still
+  // `ZHIMAN_API_KEY` (NOT `ZHIMAN_27B_API_KEY`) because the factory
+  // hardcodes that single name for ALL zhiman_* blocks — see the
+  // matching note in `defaultZhiman35bProviderBlock`.
   return [
     ZHIMAN_27B_PROVIDER_HEADER,
     'name = "zhiman_27b"',
     `# base_url = "http://your-private-deployment/v1"   # FILL IN: private vLLM / OpenAI-compatible endpoint`,
-    `# env_key  = "ZHIMAN_27B_API_KEY"                   # FILL IN: name of the env var holding your API key`,
+    `# env_key  = "ZHIMAN_API_KEY"                      # FILL IN: name of the env var holding your API key`,
     `wire_api = "${DEFAULT_PROVIDER_WIRE_API}"`,
+    `# http_headers = { "X-Custom" = "value" }          # optional: literal per-request headers`,
+    `# env_http_headers = { "X-Tenant" = "TENANT_ID_ENV" }  # optional: headers from env-var indirection`,
+    `# query_params = { "api-version" = "2024-10" }     # optional: extra URL query params`,
+    `# request_max_retries = 4                          # optional: HTTP retry budget`,
+    `# stream_max_retries = 5                           # optional: SSE stream reconnect retry budget`,
+    `# stream_idle_timeout_ms = 300000                   # optional: idle cap (default 300_000 = 5 min)`,
+    `# request_body_extras = { chat_template_kwargs = { enable_thinking = false } }  # optional: qwen thinking toggle (private vLLM)`,
+    `# experimental_bearer_token = ""                   # discouraged: prefer env_key`,
   ].join("\n");
 }
 
@@ -653,11 +739,25 @@ function defaultDashscopeProviderBlock() {
   // either uncomment this line and set DASHSCOPE_API_KEY, or leave
   // it commented and set DASHSCOPE_API_KEY as a regular env var, both
   // paths converge.
+  //
+  // DashScope-specific note: the top-level `enable_thinking` flag is
+  // the canonical thinking toggle (NOT the nested `chat_template_kwargs`
+  // form, which DashScope rejects with HTTP 400). Use
+  // `request_body_extras` to inject it if you do not want to set
+  // `model_reasoning_effort` at the root.
   return [
     DASHSCOPE_PROVIDER_HEADER,
     'name = "DashScope"',
     `# base_url = "https://dashscope.aliyuncs.com/compatible-mode/v1"   # FILL IN: DashScope OpenAI-compatible endpoint`,
     `# env_key  = "DASHSCOPE_API_KEY"                                   # FILL IN: name of the env var holding your DashScope API key`,
     `wire_api = "${DEFAULT_PROVIDER_WIRE_API}"`,
+    `# http_headers = { "X-Custom" = "value" }          # optional: literal per-request headers`,
+    `# env_http_headers = { "X-Tenant" = "TENANT_ID_ENV" }  # optional: headers from env-var indirection`,
+    `# query_params = { "api-version" = "2024-10" }     # optional: extra URL query params`,
+    `# request_max_retries = 4                          # optional: HTTP retry budget`,
+    `# stream_max_retries = 5                           # optional: SSE stream reconnect retry budget`,
+    `# stream_idle_timeout_ms = 300000                   # optional: idle cap (default 300_000 = 5 min)`,
+    `# request_body_extras = { enable_thinking = false }   # optional: DashScope thinking toggle (top-level form)`,
+    `# experimental_bearer_token = ""                   # discouraged: prefer env_key`,
   ].join("\n");
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@zhiman_innies/innies-codex",
-  "version": "0.122.50",
+  "version": "0.122.54",
   "license": "Apache-2.0",
   "bin": {
     "innies": "bin/innies.js"
@@ -23,9 +23,9 @@
     "postinstall": "node bin/innies-init.js"
   },
   "optionalDependencies": {
-    "@zhiman_innies/innies-codex-darwin-x64": "0.122.50-darwin-x64",
-    "@zhiman_innies/innies-codex-darwin-arm64": "0.122.50-darwin-arm64",
-    "@zhiman_innies/innies-codex-win32-x64": "0.122.50-win32-x64",
-    "@zhiman_innies/innies-codex-win32-arm64": "0.122.50-win32-arm64"
+    "@zhiman_innies/innies-codex-darwin-x64": "0.122.54-darwin-x64",
+    "@zhiman_innies/innies-codex-darwin-arm64": "0.122.54-darwin-arm64",
+    "@zhiman_innies/innies-codex-win32-x64": "0.122.54-win32-x64",
+    "@zhiman_innies/innies-codex-win32-arm64": "0.122.54-win32-arm64"
   }
 }