ho ho ho tool call service uses full subagen's chat history to make tool call

Author: sidwan02

packages/core/src/agents/executor.ts

@@ -65,7 +65,7 @@ import { zodToJsonSchema } from 'zod-to-json-schema';
 import { debugLogger } from '../utils/debugLogger.js';
 import type { Part as OllamaPart } from '../core/ollamaChat.js';
 // import { extractValidJson } from '../utils/json.js';
-import { stripJsonMarkdown } from '../utils/json.js';
+
 import * as fs from 'node:fs/promises';
 import type { AnsiOutput } from '../utils/terminalSerializer.js';
 import {

packages/core/src/services/toolCallService.ts

@@ -108,26 +108,36 @@ export class ToolCallService {
       .join(' ');
 
     const systemPrompt = `You are a tool call generator. Your task is to generate a valid tool call for the tool named "${toolName}".
-You will be provided with the schema of the tool and the last user message.
+You will be provided with the schema of the tool and the chat history.
 Generate a tool call that best helps with the user's request.
 The tool schema is:
 ${formattedSchema}
 
 Respond with only the JSON for the tool call. Do not include any other text or markdown.
+
+The 
 `;
 
+    const reminder = `Remember! 
+You must look at the preceding chat history to understand the user's objective and current step in the process.
+Only respond with the JSON for the tool call. Do not include any other text or markdown.`;
+
+    // TODO: add a user message which actually has the formatted schema and a reiteration of the user's objective and asks what command must be called. that will replace text: ''. maybe it's fine to have the schema and tool name in the system prompt.
+
     const tempChat = new OllamaChat(
       modelConfig as OllamaModelConfig,
       systemPrompt,
-      [],
+      // [],
+      chatHistory,
       {
         query: '',
         systemPrompt,
       },
     );
 
     const responseStream = await tempChat.sendMessageStream(modelConfig.model, {
-      message: [{ text: lastMessageText }],
+      // message: [{ text: lastMessageText }],
+      message: [{ text: reminder }],
     });
 
     let textResponse = '';

understand.md

@@ -248,11 +248,20 @@ Using the codebase investigator: tell me how the ollama inference works.
    environment either due to prompt forgetting)
 1. Gemma often forgets which files it's already looked at.
 
-1. Look at chat_history.txt. despite having a summary it still sees status 0 and
-   decidesit doesn't have the full information.
+<!--  -->
 
-1. the dircetive swapping between tool choice and tool call is a little messy.
-   just verify that later in executor.ts
+1. chat_history maybe shouldn't have the model's tool choice. actually this
+   could be fine.
+1. Tool call service should take in the full chat history since if the model's
+   tool choice is provided then the last mesage is just the tool choice action
+   by the main subagent loop.
+1. When tool service makes a tool call, that should be reflected in the UI. In
+   general, subagent services should be reflected in the UI.
+
+<!-- For routing -->
+
+If ollama hasn't started then this error comes out from routing:
+http://gpaste/5701046179594240
 
 # Litellm setup