@purista/harness-openai 1.2.1 → 1.2.3

package/dist/index.js

@@ -67,33 +67,31 @@ class OpenAiModelProvider extends BaseModelProvider {
         req.signal.throwIfAborted();
         const stream = await createChatCompletion(this.client, req, true);
         let usage = { inputTokens: 0, outputTokens: 0, totalTokens: 0 };
+        let finishReason = 'stop';
+        const toolState = new Map();
         for await (const chunk of stream) {
             req.signal.throwIfAborted();
-            const choice = chunk.choices[0];
+            // The usage chunk arrives with an empty choices array, so read it first.
+            if (chunk.usage) {
+                usage = toUsage(chunk.usage.prompt_tokens, chunk.usage.completion_tokens);
+            }
+            const choice = chunk.choices?.[0];
             if (!choice)
                 continue;
             if (choice.delta?.content) {
                 yield { kind: 'delta', text: choice.delta.content };
             }
             if (choice.delta?.tool_calls) {
-                for (const call of choice.delta.tool_calls) {
-                    if (!call.function?.name || !call.id)
-                        continue;
-                    yield {
-                        kind: 'tool_call',
-                        call: {
-                            id: call.id,
-                            name: call.function.name,
-                            arguments: parseToolArgs(call.function.arguments, req, 'textStream')
-                        }
-                    };
-                }
+                accumulateToolCallDeltas(toolState, choice.delta.tool_calls);
             }
-            if (chunk.usage) {
-                usage = toUsage(chunk.usage.prompt_tokens, chunk.usage.completion_tokens);
+            if (choice.finish_reason) {
+                finishReason = toFinishReason(choice.finish_reason);
             }
         }
-        yield { kind: 'finish', usage, finishReason: 'stop' };
+        for (const call of finalizeStreamToolCalls(toolState, req, 'textStream')) {
+            yield { kind: 'tool_call', call };
+        }
+        yield { kind: 'finish', usage, finishReason };
     }
     async doObject(req) {
         req.signal.throwIfAborted();
@@ -112,10 +110,15 @@ class OpenAiModelProvider extends BaseModelProvider {
         req.signal.throwIfAborted();
         let partial = '';
         let usage = { inputTokens: 0, outputTokens: 0, totalTokens: 0 };
+        let finishReason = 'stop';
+        const toolState = new Map();
         const stream = await createChatCompletion(this.client, req, true);
         for await (const chunk of stream) {
             req.signal.throwIfAborted();
-            const choice = chunk.choices[0];
+            if (chunk.usage) {
+                usage = toUsage(chunk.usage.prompt_tokens, chunk.usage.completion_tokens);
+            }
+            const choice = chunk.choices?.[0];
             if (!choice)
                 continue;
             if (choice.delta?.content) {
@@ -123,25 +126,17 @@ class OpenAiModelProvider extends BaseModelProvider {
                 yield { kind: 'partial', partial: safePartialJson(partial) };
             }
             if (choice.delta?.tool_calls) {
-                for (const call of choice.delta.tool_calls) {
-                    if (!call.function?.name || !call.id)
-                        continue;
-                    yield {
-                        kind: 'tool_call',
-                        call: {
-                            id: call.id,
-                            name: call.function.name,
-                            arguments: parseToolArgs(call.function.arguments, req, 'objectStream')
-                        }
-                    };
-                }
+                accumulateToolCallDeltas(toolState, choice.delta.tool_calls);
             }
-            if (chunk.usage) {
-                usage = toUsage(chunk.usage.prompt_tokens, chunk.usage.completion_tokens);
+            if (choice.finish_reason) {
+                finishReason = toFinishReason(choice.finish_reason);
             }
         }
+        for (const call of finalizeStreamToolCalls(toolState, req, 'objectStream')) {
+            yield { kind: 'tool_call', call };
+        }
         const object = parseJson(partial || '{}', req, 'objectStream');
-        yield { kind: 'finish', object, usage, finishReason: 'stop' };
+        yield { kind: 'finish', object, usage, finishReason };
     }
     async doEmbed(req) {
         req.signal.throwIfAborted();
@@ -200,11 +195,14 @@ async function createChatCompletion(client, req, stream) {
         model: req.model,
         messages,
         stream,
+        // OpenAI only emits a usage chunk during streaming when this is set.
+        ...(stream ? { stream_options: { include_usage: true } } : {}),
         tools: toTools(req.tools),
         temperature: req.call?.temperature ?? req.defaults?.temperature,
         max_tokens: req.call?.maxTokens ?? req.defaults?.maxTokens,
         top_p: req.call?.topP ?? req.defaults?.topP,
         stop: req.call?.stopSequences ?? req.defaults?.stopSequences,
+        ...(req.tools && (req.call?.parallelToolCalls ?? req.defaults?.parallelToolCalls) !== undefined ? { parallel_tool_calls: req.call?.parallelToolCalls ?? req.defaults?.parallelToolCalls } : {}),
         response_format: toResponseFormat(req),
         ...bodyOptions
     }, { ...requestOptions, signal: req.signal });
@@ -287,6 +285,29 @@ function toTools(tools) {
         }
     }));
 }
+function accumulateToolCallDeltas(state, deltas) {
+    for (const delta of deltas) {
+        const index = typeof delta?.index === 'number' ? delta.index : 0;
+        const existing = state.get(index) ?? { args: '' };
+        if (delta?.id)
+            existing.id = String(delta.id);
+        if (delta?.function?.name)
+            existing.name = String(delta.function.name);
+        if (typeof delta?.function?.arguments === 'string')
+            existing.args += delta.function.arguments;
+        state.set(index, existing);
+    }
+}
+function finalizeStreamToolCalls(state, req, method) {
+    return [...state.entries()]
+        .sort((a, b) => a[0] - b[0])
+        .filter(([, call]) => call.id && call.name)
+        .map(([, call]) => ({
+        id: call.id,
+        name: call.name,
+        arguments: parseToolArgs(call.args || undefined, req, method)
+    }));
+}
 function parseToolArgs(argumentsText, req, method) {
     if (!argumentsText)
         return {};

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@purista/harness-openai",
-  "version": "1.2.1",
+  "version": "1.2.3",
   "description": "OpenAI model provider adapter for @purista/harness.",
   "type": "module",
   "main": "./dist/index.js",