@axlsdk/axl 0.22.0 → 0.22.1

package/README.md

@@ -67,7 +67,7 @@ agent({ model: 'deepseek:deepseek-reasoner' });           // DEEPSEEK_API_KEY
 agent({ model: 'ollama:llama3' });                        // local — no key, $0
 ```
 
-Presets: `openrouter`, `azure`, `xai`, `deepseek`, `mistral`, `groq`, `bedrock`, and self-hosted `ollama` / `vllm` / `lmstudio` / `llamacpp` / `sglang`. The unified `effort` knob and per-call cost tracking work across them. Build your own by cloning a `ProviderProfile`. See [docs/providers.md](../../docs/providers.md#openai-compatible-providers--presets).
+Presets: `openrouter`, `azure`, `xai`, `deepseek`, `mistral`, `groq`, `bedrock`, and self-hosted `ollama` / `vllm` / `lmstudio` / `llamacpp` / `sglang`. The unified `effort` knob and per-call cost tracking work across them. OpenRouter passes image input through to the selected model/route, but Axl does not claim every catalog model supports images, tools, or transcription. Build your own by cloning a `ProviderProfile`; see [provider details](../../docs/providers.md#openai-compatible-providers--presets) and [multimodal input](../../docs/multimodal-input.md).
 
 Each provider also accepts an opt-in `rateLimit` (`{ maxConcurrent?, minIntervalMs?, acquireTimeoutMs? }`) for proactive client-side pacing on top of the automatic 429/503/529 backoff — useful when a large fan-out (e.g. an eval) shares one API key. It caps in-flight request concurrency (not token throughput) for that provider's chat calls. See [Providers → Rate limiting](../../docs/providers.md#rate-limiting-opt-in).
 
@@ -792,10 +792,12 @@ See [docs/providers.md](../../docs/providers.md) for the full model list includi
 
 Use `ctx.transcribe()` for finite recordings, then explicitly pass
 `Transcript.text` to an agent if analysis is needed. It is separate from chat
-history and chat providers. B1 supports exact OpenAI, Gemini, and OpenRouter
-transcription URIs; see [Multimodal input](../../docs/multimodal-input.md) for
-the 25 MiB inline bound, source types, options, transcript redaction, temporary
-Gemini Files cleanup, and live-verification commands.
+history and chat providers. OpenAI and Gemini use their documented model
+contracts; any nonblank `openrouter-transcription:<vendor/model>` URI can
+dispatch supported bytes/base64 to OpenRouter, where the selected model/route
+decides endpoint compatibility. See [Multimodal input](../../docs/multimodal-input.md)
+for the 25 MiB inline bound, source types, options, transcript redaction,
+temporary Gemini Files cleanup, and live-verification commands.
 
 ## License
 

package/dist/index.cjs

@@ -1277,13 +1277,12 @@ function unsafeTransportError(surface, remediation, exposedData) {
 }
 
 // src/providers/openai-compatible.ts
-var OPENROUTER_IMAGE_MODEL = "openai/gpt-4o-mini";
 function compatibleImageBase64(source) {
   return source.type === "base64" ? source.data : Buffer.from(source.data.buffer, source.data.byteOffset, source.data.byteLength).toString(
     "base64"
   );
 }
-function openRouterImageParts(parts) {
+function openRouterImageParts(parts, model) {
   const content = [];
   for (const part of parts) {
     if (part.type === "text") {
@@ -1294,7 +1293,7 @@ function openRouterImageParts(parts) {
     if (source.type === "provider-file") {
       throw new UnsupportedModelInputError({
         provider: "openrouter",
-        model: OPENROUTER_IMAGE_MODEL,
+        model,
         modality: "image",
         source: "provider-file"
       });
@@ -1425,7 +1424,7 @@ function normalizeMessageContent(content) {
 var OpenAICompatibleProvider = class {
   name;
   inputCapabilities(model) {
-    return this.name === "openrouter" && model === OPENROUTER_IMAGE_MODEL ? { image: { sources: ["url", "bytes", "base64"] } } : {};
+    return this.name === "openrouter" && model.trim().length > 0 ? { image: { sources: ["url", "bytes", "base64"] } } : {};
   }
   validateInput(request) {
     const effectiveModel = typeof request.providerOptions?.model === "string" ? request.providerOptions.model : request.model;
@@ -1438,13 +1437,12 @@ var OpenAICompatibleProvider = class {
         ...feature ? { feature } : {}
       });
     };
-    if (this.name !== "openrouter" || effectiveModel !== OPENROUTER_IMAGE_MODEL) {
+    if (this.name !== "openrouter" || effectiveModel.trim().length === 0) {
       fail(void 0, "image input for this model");
     }
     if (request.providerOptions && "messages" in request.providerOptions) {
       fail(void 0, "raw messages providerOptions");
     }
-    if (request.hasTools) fail(void 0, "tools with image input");
     for (const message of request.history) {
       if (!Array.isArray(message.content)) continue;
       if (message.role !== "user") fail(void 0, "rich non-user history");
@@ -1632,7 +1630,7 @@ var OpenAICompatibleProvider = class {
   formatMessage(msg, model) {
     const out = {
       role: this.profile.roleFor ? this.profile.roleFor(msg.role, model) : msg.role,
-      content: Array.isArray(msg.content) && this.name === "openrouter" && model === OPENROUTER_IMAGE_MODEL ? openRouterImageParts(msg.content) : msg.content
+      content: Array.isArray(msg.content) && this.name === "openrouter" ? openRouterImageParts(msg.content, model) : msg.content
     };
     if (msg.name && (this.profile.capabilities?.emitsMessageName ?? true)) out.name = msg.name;
     if (msg.tool_calls) out.tool_calls = msg.tool_calls;
@@ -5783,7 +5781,6 @@ function normalizeTranscriptionAccounting(usageValue, pricingValue) {
 // src/providers/transcription-utils.ts
 var OPENAI_TRANSCRIPTION_MODEL = "gpt-transcribe";
 var GEMINI_TRANSCRIPTION_MODEL = "gemini-3.5-transcribe";
-var OPENROUTER_TRANSCRIPTION_MODEL = "openai/whisper-1";
 function requireExactTranscriptionModel(provider, model, expected) {
   if (model !== expected) {
     throw new UnsupportedTranscriptionInputError({
@@ -6434,10 +6431,16 @@ var OpenRouterTranscriptionProvider = class {
     this.governor = options.rateLimit ? new RateLimiter(options.rateLimit) : void 0;
   }
   capabilities(model) {
-    return model === OPENROUTER_TRANSCRIPTION_MODEL ? { sources: ["bytes", "base64"] } : void 0;
+    return model.trim().length > 0 ? { sources: ["bytes", "base64"] } : void 0;
   }
   async transcribe(request) {
-    requireExactTranscriptionModel(this.name, request.model, OPENROUTER_TRANSCRIPTION_MODEL);
+    if (request.model.trim().length === 0) {
+      throw new UnsupportedTranscriptionInputError({
+        provider: this.name,
+        model: request.model,
+        feature: "this transcription model"
+      });
+    }
     requireNoProviderFile(request.audio, this.name, request.model);
     if (request.audio.type === "provider-file")
       throw new UnsupportedTranscriptionInputError({
@@ -6577,7 +6580,7 @@ var TranscriptionProviderRegistry = class {
   }
   resolve(uri, config = {}) {
     const colon = typeof uri === "string" ? uri.indexOf(":") : -1;
-    if (colon <= 0 || colon === uri.length - 1 || uri.indexOf(":", colon + 1) !== -1) {
+    if (colon <= 0 || colon === uri.length - 1) {
       throw new UnsupportedTranscriptionInputError({
         provider: "unknown",
         model: "",