mohdel 0.122.0 → 0.124.0

package/README.md

@@ -154,7 +154,9 @@ const result = await mo.use('anthropic/claude-sonnet-4-6').answer('Hello')
 console.log(result.output, result.cost)
 ```
 
-No subprocess, no setup beyond your API key. Right for CLI tools (`mo ask`), scripts, tests, and single-process services — which is most projects.
+No subprocess, no setup beyond your API key. Right for CLI tools (`mo ask`), scripts, tests, and single-process services — which is most projects. Pass an `AbortSignal` as `answer(prompt, { signal })` to cancel in flight.
+
+The factory reads the catalog from `~/.config/mohdel/curated.json` and keys from the environment, the same defaults a gate session starts from. `mohdel({ models })` replaces the catalog for the process, factory and session runtime alike, the way `set_catalog` replaces it in a gate session. `mohdel({ configurations: { openai: { apiKey } } })` overrides the key per provider.
 
 ### Client — cross-process (the production gateway)
 
@@ -233,9 +235,26 @@ With no session-bin configured, thin-gate runs in demo mode: `POST /v1/call` ret
 
 The client snippet under [Library Usage](#library-usage) above is the full surface: `call(envelope, { socketPath, signal? })` returns an async iterable of events. Pass an `AbortSignal` to cancel in flight; thin-gate forwards a cancel control message to the session and reuses it on the pool. The envelope is the flat `answer(prompt, options)` surface plus transport metadata (`callId`, `authId`, `auth.key`, optional `traceparent`); see [`js/core/envelope.js`](js/core/envelope.js) for the full field list.
 
+### Other languages
+
+The gate's HTTP surface is specified in [PROTOCOL.md §10](PROTOCOL.md#10-gate-http-surface-clients); `test/conformance/` holds the fixtures a client round-trips. Reference clients live under `clients/`:
+
+- **Lua** — [`clients/lua`](clients/lua): Lua 5.1+ / LuaJIT, transports over LuaSocket or `curl`. `cd clients/lua && luarocks make`.
+- **Gleam** — [`clients/gleam`](clients/gleam): Erlang target, typed events and results, `gen_tcp` over the unix socket. Path dependency until it is on Hex.
+- **Rust** — [`clients/rust`](clients/rust): async (tokio), uses the gate's own wire types from the `mohdel-protocol` crate. Path dependency until it is on crates.io.
+- **OCaml** — [`clients/ocaml`](clients/ocaml): synchronous, stdlib `Unix` transport, `yojson`; OCaml 4.14+. `opam pin` until it is on opam.
+
+```lua
+local mohdel = require('mohdel')
+local c = mohdel.connect{ socket = '/tmp/mohdel-data.sock' }
+for ev in c:call(envelope):events() do
+  if ev.type == 'delta' then io.write(ev.delta.delta) end
+end
+```
+
 ### Canonical types (frozen wire contract)
 
-Wire format is JSON over NDJSON frames, camelCase. Types are defined in `js/core/` (JSDoc) and mirrored in `rust/thin-gate/src/protocol.rs` (serde). Cross-language conformance tests enforce round-trip fidelity. The session-side protocol (envelopes in, events out, cancel control messages) is specified in [PROTOCOL.md](PROTOCOL.md) — read that to implement a session in another language.
+Wire format is JSON over NDJSON frames, camelCase. Types are defined in `js/core/` (JSDoc) and mirrored in `rust/protocol/src/protocol.rs` (serde, the `mohdel-protocol` crate). Cross-language conformance tests enforce round-trip fidelity. The session-side protocol (envelopes in, events out, cancel control messages) is specified in [PROTOCOL.md](PROTOCOL.md) — read that to implement a session in another language.
 
 - **`CallEnvelope`** — flat `answer()` options plus transport metadata: `callId`, `authId`, `auth.key`, `traceparent?`, `baggage?`, `provider`, `model`, `prompt`, `outputBudget?`, `outputType?`, `outputStyle?`, `outputEffort?`, `images?`, `videos?`, `cache?`, `tools?`, `toolChoice?`, `parallelToolCalls?`, `identifier?`.
 - **`Event`** — three-variant union discriminated on `type`:

package/js/factory/bridge.js

@@ -35,7 +35,6 @@ import { createRealtimeDeltaBuffer } from '../../src/lib/utils.js'
  *   Hook that returns `{rpmLimit, tpmLimit}` for a given provider —
  *   lets the factory keep its own `providersConfig` as source of
  *   truth instead of the module-level `providers.json` loader.
- * @property {AbortSignal} [signal]
  */
 
 /**
@@ -58,6 +57,7 @@ import { createRealtimeDeltaBuffer } from '../../src/lib/utils.js'
  * | `providerOrder` / `providerAllow` / `providerDeny`| envelope.providerOptions.openrouter                  |
  * | `traceparent` / `baggage`                         | envelope transport metadata                          |
  * | `callId` / `authId`                               | envelope transport metadata                          |
+ * | `signal`                                          | not on the wire — `run()` deps, aborts the call      |
  * | `configuration.apiKey`                            | envelope.auth.key                                    |
  * | **`parentSpan`**                                  | **dropped** — use `traceparent` instead              |
  * | **`maybeThrowHandler`**                           | **dropped** — no factory-side validation hook        |
@@ -90,7 +90,7 @@ export async function runAnswer ({ provider, model, modelKey, configuration, pro
 
   let terminal
   try {
-    for await (const ev of run(envelope, deps)) {
+    for await (const ev of run(envelope, { ...deps, signal: options.signal })) {
       if (ev.type === 'delta' && ev.delta) {
         if (deltaBuffer) deltaBuffer.push(ev.delta.type, ev.delta.delta)
       } else if (ev.type === 'done' || ev.type === 'error') {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mohdel",
-  "version": "0.122.0",
+  "version": "0.124.0",
   "license": "MIT",
   "author": {
     "name": "Christophe Le Bars",
@@ -108,16 +108,16 @@
     "@opentelemetry/exporter-trace-otlp-grpc": "^0.222.0",
     "@opentelemetry/sdk-node": "^0.222.0",
     "chalk": "^6.0.0",
-    "mohdel-thin-gate-linux-x64-gnu": "0.122.0"
+    "mohdel-thin-gate-linux-x64-gnu": "0.124.0"
   },
   "dependencies": {
-    "@anthropic-ai/sdk": "^0.122.0",
+    "@anthropic-ai/sdk": "^0.123.0",
     "@cerebras/cerebras_cloud_sdk": "^1.91.0",
     "@google/genai": "^2.20.0",
     "@opentelemetry/api": "^1.9.1",
     "env-paths": "^4.0.0",
     "groq-sdk": "^1.6.0",
-    "openai": "^7.8.0",
+    "openai": "^7.9.0",
     "undici": "^7.29.0"
   },
   "lint-staged": {

package/src/cli/ask.js

@@ -13,7 +13,7 @@ export const hintsForError = (err, modelId) => {
   const provider = modelId.includes('/') ? modelId.split('/')[0] : null
   const hints = []
 
-  if (/not found in curated models/i.test(both)) {
+  if (/not found in catalog/i.test(both)) {
     if (provider) {
       hints.push(`→ run:  mo curate ${provider}        # add upstream models from this provider`)
       hints.push(`→ or:   mo model add ${modelId}      # add this one manually`)

package/src/lib/common.js

@@ -112,12 +112,11 @@ const createFileOperation = (filePath, defaultValue = {}, operationType) => {
   const loadHandler = async () => {
     let loadedData
     try {
-      if (!existsSync(CONFIG_DIR)) {
-        await mkdir(CONFIG_DIR, { recursive: true })
-      }
-
       if (!existsSync(filePath)) {
         if (defaultValue && Object.keys(defaultValue).length > 0) {
+          if (!existsSync(CONFIG_DIR)) {
+            await mkdir(CONFIG_DIR, { recursive: true })
+          }
           await writeFile(filePath, JSON.stringify(defaultValue, null, 2))
           loadedData = JSON.parse(JSON.stringify(defaultValue))
         } else {

package/src/lib/curated-cache.js

@@ -69,6 +69,11 @@ const rebuildAliasMap = () => {
 
 export const loadCuratedCache = ensureCuratedCache
 
+export const setCuratedCache = (table) => {
+  curatedCache = { ...table }
+  aliasMapCache = buildAliasMap(curatedCache)
+}
+
 export const getCuratedCacheSnapshot = () => curatedCache
 
 export const getAliasMapSnapshot = () => aliasMapCache

package/src/lib/index.js

@@ -7,6 +7,7 @@ import providers from './providers.js'
 import { getAPIKey, loadDefaultEnv, getProvidersConfig, saveProvidersConfig, setLogger as setCommonLogger } from './common.js'
 import {
   loadCuratedCache,
+  setCuratedCache,
   getCuratedCacheSnapshot,
   expandModelAliasSync,
   suggestModels,
@@ -14,6 +15,7 @@ import {
 } from './curated-cache.js'
 import { createRateLimiter } from '../../js/session/_rate_limiter.js'
 import { createCooldownTracker } from '../../js/session/_cooldown.js'
+import { setCatalog } from '../../js/session/adapters/_catalog.js'
 import { runAnswer, runAnswerImage, runAnswerTranscription } from '../../js/factory/bridge.js'
 import { startSpan, endSpanOk, endSpanError } from './tracing.js'
 import { isValidTag } from './schema.js'
@@ -200,21 +202,27 @@ export { buildHandlers as _buildHandlersForTests }
  * @param {Function} [opts.onFailure]           — fired after every failed answer call
  * @param {number} [opts.cooldownThreshold=3]   — consecutive provider failures before cooldown
  * @param {number} [opts.cooldownDuration=60000] — cooldown duration in ms
- * @param {object} [opts.models]                — model catalog (library mode — skips disk init)
- * @param {object} [opts.configurations]        — provider configurations (library mode)
+ * @param {object} [opts.models]                — catalog `{ [provider/model]: spec }`. Replaces
+ *                                                `~/.config/mohdel/curated.json` for this process,
+ *                                                factory and session runtime alike.
+ * @param {object} [opts.configurations]        — `{ [provider]: { apiKey } }`. Overrides the env
+ *                                                key per provider; env for the rest.
  */
 const mohdel = async ({ logger, verbosity: verbosityOpt, onSuccess, onFailure, cooldownThreshold, cooldownDuration, models, configurations } = {}) => {
-  // When consumer provides models + configurations, skip disk-based init (library mode).
-  // Otherwise load from ~/.config/mohdel/ (CLI / standalone mode).
-  const libraryMode = !!(models && configurations)
-
-  if (!libraryMode) {
-    loadDefaultEnv()
+  loadDefaultEnv()
+
+  // The session runtime keeps its own catalog cache (the seam thin-gate's
+  // `set_catalog` uses) and would otherwise lazy-load curated.json, so a
+  // replacement has to reach both caches.
+  if (models) {
+    setCuratedCache(models)
+    setCatalog(models)
+  } else {
     await loadCuratedCache()
   }
 
   // Provider config: user-specific rate limits per provider (~/.config/mohdel/providers.json)
-  const providersConfig = libraryMode ? {} : await getProvidersConfig()
+  const providersConfig = await getProvidersConfig()
 
   // Resolve verbosity tier once at init. Factory opt > env var > default.
   // Captured in the answer() closure below to gate per-call log lines.
@@ -248,7 +256,7 @@ const mohdel = async ({ logger, verbosity: verbosityOpt, onSuccess, onFailure, c
     get: (target, prop) => {
       if (prop === 'list') {
         return (tag) => { // Sync
-          const catalog = libraryMode ? models : getCuratedCacheSnapshot()
+          const catalog = getCuratedCacheSnapshot()
           let modelEntries = Object.entries(catalog)
             .filter(([, metadata]) => !metadata.deprecated)
 
@@ -267,7 +275,7 @@ const mohdel = async ({ logger, verbosity: verbosityOpt, onSuccess, onFailure, c
 
       if (prop === 'use') {
         return (modelId) => { // Sync
-          const catalog = libraryMode ? models : getCuratedCacheSnapshot()
+          const catalog = getCuratedCacheSnapshot()
 
           // Parse optional :outputEffort suffix (e.g. "claude-opus:max").
           // Effort levels vary per spec (`thinkingEffortLevels` keys —
@@ -284,7 +292,7 @@ const mohdel = async ({ logger, verbosity: verbosityOpt, onSuccess, onFailure, c
           // Fallback specs are excluded from that test on purpose:
           // providers that synthesize them resolve any string, which
           // would disable suffix parsing for them entirely.
-          const exactId = libraryMode ? modelId : expandModelAliasSync(modelId)
+          const exactId = expandModelAliasSync(modelId)
           const isCuratedId = !!catalog[exactId]
 
           // `@speed` is parsed off first so the two suffixes are read
@@ -299,7 +307,7 @@ const mohdel = async ({ logger, verbosity: verbosityOpt, onSuccess, onFailure, c
             // resolves the same as `x@fast`.
             const colon = base.lastIndexOf(':')
             const probe = colon > 0 ? base.slice(0, colon) : base
-            const probeResolved = libraryMode ? probe : expandModelAliasSync(probe)
+            const probeResolved = expandModelAliasSync(probe)
             if (catalog[probeResolved] || createFallbackModelSpec(probeResolved)) {
               aliasSpeed = candidate
               modelId = base
@@ -311,7 +319,7 @@ const mohdel = async ({ logger, verbosity: verbosityOpt, onSuccess, onFailure, c
           if (colonIdx > 0) {
             const candidate = modelId.slice(colonIdx + 1)
             const base = modelId.slice(0, colonIdx)
-            const baseResolved = libraryMode ? base : expandModelAliasSync(base)
+            const baseResolved = expandModelAliasSync(base)
             const baseSpec = catalog[baseResolved] || createFallbackModelSpec(baseResolved)
             if (baseSpec) {
               aliasOutputEffort = candidate
@@ -319,17 +327,14 @@ const mohdel = async ({ logger, verbosity: verbosityOpt, onSuccess, onFailure, c
             }
           }
 
-          let resolvedModelId = libraryMode ? modelId : expandModelAliasSync(modelId)
+          let resolvedModelId = expandModelAliasSync(modelId)
           let modelSpec = catalog[resolvedModelId]
 
           if (!modelSpec) {
             modelSpec = createFallbackModelSpec(resolvedModelId)
             if (!modelSpec) {
-              if (libraryMode) {
-                throw new Error(`Model '${modelId}' not found in provided models.`)
-              }
               const suggestions = suggestModels(modelId)
-              let msg = `Model '${modelId}' not found in curated models.`
+              let msg = `Model '${modelId}' not found in catalog.`
               if (suggestions.length) {
                 msg += ' Did you mean?\n' + suggestions.map(s => `  ${s.id}  ${s.label}`).join('\n')
               }
@@ -777,7 +782,7 @@ const createModelProxy = (resolvedModelId, modelSpec, handlers, aliasOutputEffor
 
       if (prop === 'info') {
         return () => { // Sync
-          const catalog = externalConfigurations ? null : getCuratedCacheSnapshot()
+          const catalog = getCuratedCacheSnapshot()
           return catalog?.[resolvedModelId] ? { ...catalog[resolvedModelId] } : { ...modelSpec }
         }
       }