Composio powers 1000+ toolkits, tool search, context management, authentication, and a sandboxed workbench to help you build AI agents that turn intent into action.
Mirrored from https://github.com/Camo67/composio.git
Synced Nov 21, 12:15 UTC
docs(openai): align responses examples on gpt-5 and document provider types
- Use gpt-5 in the Responses API examples; gpt-4 predates the Responses API and the repo's other Responses examples use gpt-5. - Print response.output_text instead of indexing into content items, which assumes non-empty message content. - Add the OpenAIResponsesProvider type surface to the Type Definitions section, which previously only showed the chat completions provider.
jkomyno committed
Sep 15, 2026 at 16:42 UTC
cedf1a421782f6ea525ad43a293085fc3595e102
1 file changed
+46
-7
ts/docs/providers/openai.md
+46
-7
@@ -155,7 +155,7 @@ const tools = await composio.tools.get('default', {
155
});
156
157
let response = await openai.responses.create({
158
- model: 'gpt-4',
158
+ model: 'gpt-5',
159
tools,
160
input: 'Find information about the Composio SDK repository',
161
});
@@ -163,18 +163,14 @@ let response = await openai.responses.create({
163
while (response.output.some(item => item.type === 'function_call')) {
164
const toolOutputs = await composio.provider.handleToolCalls('default', response.output);
165
response = await openai.responses.create({
166
- model: 'gpt-4',
166
+ model: 'gpt-5',
167
tools,
168
previous_response_id: response.id,
169
input: toolOutputs,
170
});
171
}
172
173
-for (const item of response.output) {
174
- if (item.type === 'message' && item.content[0].type === 'output_text') {
175
- console.log(item.content[0].text);
176
- }
177
-}
173
+console.log(response.output_text);
174
```
175
176
## Modifiers with OpenAI Provider
@@ -261,4 +257,47 @@ class OpenAIProvider extends BaseNonAgenticProvider<OpenAiToolCollection, OpenAi
257
modifiers?: ExecuteToolModifiers
258
): Promise<OpenAI.ChatCompletionToolMessageParam[]>;
259
}
260
+
261
+// The tool type for the Responses API
262
+// (the Responses provider's OpenAiTool is OpenAI.Responses.FunctionTool)
263
+type ResponsesTool = OpenAI.Responses.FunctionTool;
264
+
265
+// The provider class for the Responses API
266
+// (OpenAIResponsesProvider re-exports from '@composio/openai')
267
+class OpenAIResponsesProvider extends BaseNonAgenticProvider<
268
+ ResponsesTool[],
269
+ ResponsesTool
270
+> {
271
+ readonly name = 'openai';
272
+
273
+ wrapTool(tool: Tool): ResponsesTool;
274
+ wrapTools(tools: Tool[]): ResponsesTool[];
275
+
276
+ executeToolCall(
277
+ userId: string,
278
+ tool: OpenAI.Responses.ResponseFunctionToolCall,
279
+ options?: ExecuteToolFnOptions,
280
+ modifiers?: ExecuteToolModifiers
281
+ ): Promise<string>;
282
+
283
+ handleToolCalls(
284
+ userId: string,
285
+ toolCalls: OpenAI.Responses.ResponseOutputItem[],
286
+ options?: ExecuteToolFnOptions,
287
+ modifiers?: ExecuteToolModifiers
288
+ ): Promise<OpenAI.Responses.ResponseInputItem.FunctionCallOutput[]>;
289
+
290
+ handleResponse(
291
+ userId: string,
292
+ response: OpenAI.Responses.Response,
293
+ options?: ExecuteToolFnOptions,
294
+ modifiers?: ExecuteToolModifiers
295
+ ): Promise<OpenAI.Responses.ResponseInputItem.FunctionCallOutput[]>;
296
+}
297
```
298
+
299
+Like the chat completions provider, the Responses provider also accepts a
300
+`ToolCallSession` as the execution target, and it can wrap MCP servers for the
301
+Responses API via `wrapMcpServerResponse`. See
302
+`ts/packages/providers/openai/src/OpenAIResponsesProvider.ts` for the full
303
+surface.