fix(providers): thread the tool-call id through the streaming loops too

The previous commit only reached call sites written as a single-line
three-argument call. The streaming loops are formatted across lines, so
openai-compat (Groq, DeepSeek and everything else routing through it), Anthropic
and Bedrock still omitted the id and kept falling back to a fresh token.

Both Gemini paths now pass `part.functionCall?.id` — the RAW model id, not the
`ensureToolCallId` value used for stream events. That helper allocates an
execution-local id when Gemini supplies none, and it is freshly allocated per
attempt: passing it would complete the keyed context, silencing the "could not
derive" warning, while leaving the token just as unstable. Gemini frequently
omits the id, in which case this is `undefined` and the loud fallback stands.

All 27 call sites are now covered.

Author: icecrasher321

apps/sim/providers/anthropic/streaming-tool-loop.ts

@@ -367,7 +367,8 @@ export function createAnthropicStreamingToolLoopStream(
                   const { toolParams, executionParams } = prepareToolExecution(
                     tool,
                     toolArgs,
-                    request
+                    request,
+                    toolUse.id
                   )
                   const { rawResponse, modelResponse } = await executeProviderTool(
                     toolName,

apps/sim/providers/bedrock/streaming-tool-loop.ts

@@ -413,7 +413,8 @@ export function createBedrockStreamingToolLoopStream(
                 const { toolParams, executionParams } = prepareToolExecution(
                   tool,
                   toolArgs,
-                  request
+                  request,
+                  toolUse.toolUseId
                 )
                 const { rawResponse, modelResponse } = await executeProviderTool(
                   toolName,

apps/sim/providers/gemini/core.ts

@@ -133,14 +133,18 @@ async function executeToolCallsBatch(
       }
 
       /*
-       * No tool-call id is passed: Gemini's function-call parts carry no
-       * model-supplied identifier, and the streaming loop has to synthesize a
-       * local one. A positional index would not survive the model re-emitting
-       * the call, so a `keyed` tool invoked through a Gemini agent falls back to
-       * a fresh token and logs the missing identity rather than deriving one
-       * that only looks stable.
+       * The RAW model id, not a synthesized one. Gemini often omits an id on a
+       * function-call part, in which case this is `undefined` and the keyed
+       * helper falls back loudly rather than deriving a token from something —
+       * a positional index, an execution-local id — that would look stable and
+       * not be.
        */
-      const { toolParams, executionParams } = prepareToolExecution(tool, args, request)
+      const { toolParams, executionParams } = prepareToolExecution(
+        tool,
+        args,
+        request,
+        part.functionCall?.id
+      )
       const { rawResponse, modelResponse } = await executeProviderTool(toolName, executionParams, {
         signal: request.abortSignal,
       })

apps/sim/providers/gemini/streaming-tool-loop.ts

@@ -460,10 +460,21 @@ export function createGeminiStreamingToolLoopStream(
                     return value
                   }
 
+                  /*
+                   * The RAW model id, not the `ensureToolCallId` value used for
+                   * stream events: that helper falls back to an
+                   * execution-local id when Gemini supplies none, which is
+                   * freshly allocated per attempt. Passing it would complete the
+                   * keyed context — silencing the "could not derive" warning —
+                   * while leaving the token unstable, which is worse than the
+                   * loud fallback. Gemini often omits the id entirely, in which
+                   * case this is `undefined` and the fallback stands.
+                   */
                   const { toolParams, executionParams } = prepareToolExecution(
                     tool,
                     toolArgs,
-                    request
+                    request,
+                    part.functionCall?.id
                   )
                   const { rawResponse, modelResponse } = await executeProviderTool(
                     toolName,

apps/sim/providers/openai-compat/streaming-tool-loop.ts

@@ -426,7 +426,8 @@ export function createOpenAICompatStreamingToolLoopStream(
                 const { toolParams, executionParams } = prepareToolExecution(
                   tool,
                   toolArgs,
-                  request
+                  request,
+                  tc.id
                 )
                 const { rawResponse, modelResponse } = await executeProviderTool(
                   toolName,