@gaberrb/polypus 0.4.12 → 0.4.13

package/dist/index.js

@@ -910,15 +910,33 @@ var OpenAICompatibleProvider = class {
         parameters: t2.parameters
       }
     }));
+    const base = {
+      model: this.model,
+      messages,
+      ...tools && tools.length > 0 ? { tools } : {},
+      temperature: req.params?.temperature,
+      // Generous default so large files aren't truncated mid tool-call.
+      max_tokens: req.params?.maxTokens ?? 8192
+    };
+    if (req.onDelta) {
+      const stream = await this.client.chat.completions.create(
+        { ...base, stream: true, stream_options: { include_usage: true } },
+        { signal: req.signal }
+      );
+      const agg = await aggregateStream(stream, req.onDelta);
+      return {
+        content: agg.content,
+        toolCalls: agg.toolCalls.map((tc, i) => ({
+          id: tc.id || `call_${i}`,
+          name: tc.name,
+          arguments: safeParseArgs(tc.arguments)
+        })),
+        finishReason: agg.finishReason || "stop",
+        usage: agg.usage
+      };
+    }
     const completion = await this.client.chat.completions.create(
-      {
-        model: this.model,
-        messages,
-        ...tools && tools.length > 0 ? { tools } : {},
-        temperature: req.params?.temperature,
-        // Generous default so large files aren't truncated mid tool-call.
-        max_tokens: req.params?.maxTokens ?? 8192
-      },
+      { ...base },
       { signal: req.signal }
     );
     const choice = completion.choices[0];
@@ -939,6 +957,34 @@ var OpenAICompatibleProvider = class {
     };
   }
 };
+async function aggregateStream(stream, onDelta) {
+  let content = "";
+  let finishReason = "";
+  let usage2;
+  const toolAcc = [];
+  for await (const chunk of stream) {
+    const choice = chunk.choices?.[0];
+    const delta = choice?.delta;
+    if (delta?.content) {
+      content += delta.content;
+      onDelta?.(delta.content);
+    }
+    for (const tc of delta?.tool_calls ?? []) {
+      const slot = toolAcc[tc.index] ??= { id: "", name: "", arguments: "" };
+      if (tc.id) slot.id = tc.id;
+      if (tc.function?.name) slot.name = tc.function.name;
+      if (tc.function?.arguments) slot.arguments += tc.function.arguments;
+    }
+    if (choice?.finish_reason) finishReason = choice.finish_reason;
+    if (chunk.usage) {
+      usage2 = {
+        promptTokens: chunk.usage.prompt_tokens,
+        completionTokens: chunk.usage.completion_tokens
+      };
+    }
+  }
+  return { content, finishReason, usage: usage2, toolCalls: toolAcc.filter(Boolean) };
+}
 function toOpenAIMessage(m) {
   switch (m.role) {
     case "tool":
@@ -2284,13 +2330,15 @@ async function runAgent(opts) {
         }
       }
     }
+    const wantDelta = driver.kind === "native" && typeof events?.onAssistantDelta === "function";
     let response;
     try {
       response = await agent.provider.chat({
         messages,
         tools: driver.providerTools(),
         params: opts.params,
-        signal: opts.signal
+        signal: opts.signal,
+        onDelta: wantDelta ? (chunk) => events.onAssistantDelta(chunk) : void 0
       });
     } catch (err) {
       if (opts.signal?.aborted) return { finished: false, reason: "cancelled", steps: step, messages, usage: usage2 };
@@ -4406,16 +4454,27 @@ function renderDiff(hunks) {
   }
 }
 function renderEvents(spinner3) {
+  let streamed = false;
   return {
     onStep() {
+      streamed = false;
       spinner3.start(t("ui.thinking"));
     },
     onUsage(usage2) {
       const total = usage2.promptTokens + usage2.completionTokens;
       if (total > 0) spinner3.setSuffix(t("ui.tokensShort", { total: fmtTokens(total) }));
     },
+    onAssistantDelta(chunk) {
+      spinner3.stop();
+      process.stdout.write(pc8.cyan(chunk));
+      streamed = true;
+    },
     onAssistantText(text2) {
       spinner3.stop();
+      if (streamed) {
+        process.stdout.write("\n");
+        return;
+      }
       if (text2.trim()) console.log(pc8.cyan(text2.trim()));
     },
     onToolCall(call) {