@jambonz/llm 0.5.1 → 0.5.3

package/dist/index.d.cts

@@ -31,7 +31,7 @@ interface RawCredential {
 /**
  * Convert a vendor + raw credential into an `AuthSpec` discriminated union member.
  *
- * Supported vendors: openai, deepseek, anthropic, google, vertex (alias for
+ * Supported vendors: openai, deepseek, baseten, anthropic, google, vertex (alias for
  * vertex-gemini + vertex-openai — both accept the same auth shape), vertex-gemini,
  * vertex-openai, bedrock, azure-openai, groq, huggingface.
  *

package/dist/index.d.ts

@@ -31,7 +31,7 @@ interface RawCredential {
 /**
  * Convert a vendor + raw credential into an `AuthSpec` discriminated union member.
  *
- * Supported vendors: openai, deepseek, anthropic, google, vertex (alias for
+ * Supported vendors: openai, deepseek, baseten, anthropic, google, vertex (alias for
  * vertex-gemini + vertex-openai — both accept the same auth shape), vertex-gemini,
  * vertex-openai, bedrock, azure-openai, groq, huggingface.
  *

package/dist/index.js

@@ -495,7 +495,7 @@ var OpenAIAdapter = class {
   }
   stream(req) {
     return streamFromOpenAI(this.ensureClient(), req, {
-      knownModels: openAIManifest.knownModels,
+      knownModels: this.knownModels(),
       vendorMetadataExtractor: OPENAI_METADATA_EXTRACTOR
     });
   }
@@ -506,7 +506,16 @@ var OpenAIAdapter = class {
     return appendOpenAIToolResult(history, toolCallId, result);
   }
   async listAvailableModels() {
-    return listOpenAICompatibleModels(this.ensureClient(), openAIManifest.knownModels);
+    return listOpenAICompatibleModels(this.ensureClient(), this.knownModels());
+  }
+  /**
+   * Capability table used to enrich both `stream()` (for tools/vision gating)
+   * and `listAvailableModels()` (for filling capability flags on live entries).
+   * Subclasses backed by the same OpenAI wire (DeepSeek, Baseten, …) override
+   * this so their own manifest entries are consulted instead of OpenAI's.
+   */
+  knownModels() {
+    return openAIManifest.knownModels;
   }
   async testCredential() {
     await this.ensureClient().models.list();
@@ -617,12 +626,165 @@ var DeepSeekAdapter = class extends OpenAIAdapter {
       client
     );
   }
+  knownModels() {
+    return deepseekManifest.knownModels;
+  }
 };
 var deepseekFactory = {
   vendor: deepseekManifest.vendor,
   manifest: deepseekManifest,
   create: () => new DeepSeekAdapter()
 };
+var basetenManifest = {
+  vendor: "baseten",
+  displayName: "Baseten",
+  authKinds: [
+    {
+      kind: "apiKey",
+      displayName: "API Key",
+      fields: [
+        {
+          name: "apiKey",
+          label: "API Key",
+          type: "password",
+          required: true,
+          help: "Get an API key from https://app.baseten.co/settings/api_keys"
+        },
+        {
+          name: "baseURL",
+          label: "Base URL",
+          type: "url",
+          required: false,
+          default: "https://inference.baseten.co/v1",
+          help: "Defaults to Baseten Model APIs. Set to https://bridge.baseten.co/v1/direct to route to a dedicated deployment."
+        }
+      ]
+    }
+  ],
+  knownModels: [
+    {
+      id: "deepseek-ai/DeepSeek-V3.1",
+      displayName: "DeepSeek V3.1",
+      capabilities: {
+        streaming: true,
+        tools: true,
+        vision: false,
+        systemPrompt: true,
+        maxContextTokens: 163840
+      }
+    },
+    {
+      id: "deepseek-ai/DeepSeek-V4-Pro",
+      displayName: "DeepSeek V4 Pro",
+      capabilities: {
+        streaming: true,
+        tools: true,
+        vision: false,
+        systemPrompt: true,
+        maxContextTokens: 131072
+      }
+    },
+    {
+      id: "zai-org/GLM-4.7",
+      displayName: "GLM 4.7",
+      capabilities: {
+        streaming: true,
+        tools: true,
+        vision: false,
+        systemPrompt: true,
+        maxContextTokens: 2e5
+      }
+    },
+    {
+      id: "zai-org/GLM-5",
+      displayName: "GLM 5",
+      capabilities: {
+        streaming: true,
+        tools: true,
+        vision: false,
+        systemPrompt: true,
+        maxContextTokens: 202800
+      }
+    },
+    {
+      id: "moonshotai/Kimi-K2.5",
+      displayName: "Kimi K2.5",
+      capabilities: {
+        streaming: true,
+        tools: true,
+        vision: true,
+        systemPrompt: true,
+        maxContextTokens: 262e3
+      }
+    },
+    {
+      id: "moonshotai/Kimi-K2.6",
+      displayName: "Kimi K2.6",
+      capabilities: {
+        streaming: true,
+        tools: true,
+        vision: true,
+        systemPrompt: true,
+        maxContextTokens: 262e3
+      }
+    },
+    {
+      id: "MiniMaxAI/MiniMax-M2.5",
+      displayName: "MiniMax M2.5",
+      capabilities: {
+        streaming: true,
+        tools: true,
+        vision: false,
+        systemPrompt: true,
+        maxContextTokens: 204e3
+      }
+    },
+    {
+      id: "nvidia/Nemotron-120B-A12B",
+      displayName: "Nemotron Super",
+      capabilities: {
+        streaming: true,
+        tools: true,
+        vision: false,
+        systemPrompt: true,
+        maxContextTokens: 202800
+      }
+    },
+    {
+      id: "openai/gpt-oss-120b",
+      displayName: "OpenAI GPT 120B",
+      capabilities: {
+        streaming: true,
+        tools: true,
+        vision: false,
+        systemPrompt: true,
+        maxContextTokens: 128072
+      }
+    }
+  ],
+  supportsModelListing: true,
+  docsUrl: "https://docs.baseten.co/inference/model-apis/overview"
+};
+var BasetenAdapter = class extends OpenAIAdapter {
+  vendor = basetenManifest.vendor;
+  init(auth, client) {
+    super.init(
+      {
+        ...auth,
+        baseURL: auth.baseURL ?? "https://inference.baseten.co/v1"
+      },
+      client
+    );
+  }
+  knownModels() {
+    return basetenManifest.knownModels;
+  }
+};
+var basetenFactory = {
+  vendor: basetenManifest.vendor,
+  manifest: basetenManifest,
+  create: () => new BasetenAdapter()
+};
 
 // src/adapters/anthropic/adapter.ts
 import Anthropic from "@anthropic-ai/sdk";
@@ -2647,6 +2809,7 @@ var groqManifest = {
 var GROQ_METADATA_EXTRACTOR = makeMetadataExtractor([
   { header: "x-request-id", key: "request_id" },
   { header: "x-groq-region", key: "region" },
+  { header: "openai-processing-ms", key: "processing_ms", numeric: true },
   { header: "x-ratelimit-remaining-requests", key: "requests_remaining", numeric: true },
   { header: "x-ratelimit-remaining-tokens", key: "tokens_remaining", numeric: true }
 ]);
@@ -2800,6 +2963,7 @@ var huggingfaceFactory = {
 // src/adapters/index.ts
 registerAdapter(openAIFactory);
 registerAdapter(deepseekFactory);
+registerAdapter(basetenFactory);
 registerAdapter(anthropicFactory);
 registerAdapter(googleFactory);
 registerAdapter(vertexGeminiFactory);
@@ -2822,6 +2986,11 @@ function normalizeAuth(vendor, raw) {
         ...requireApiKey(vendor, raw),
         baseURL: raw.api_url ?? "https://api.deepseek.com/v1"
       };
+    case "baseten":
+      return {
+        ...requireApiKey(vendor, raw),
+        baseURL: raw.api_url ?? "https://inference.baseten.co/v1"
+      };
     case "google":
       return normalizeGoogle(raw);
     case "vertex":