Enhance system prompt

README.md

@@ -11,10 +11,11 @@ conversations can be resumed with context. One example tool is included:
   returned. The VM is created when a chat session starts and reused for all
   subsequent tool calls.
 
-The application injects a system prompt on each request that instructs the
-model to chain multiple tools when required. This prompt is **not** stored in
-the chat history but is provided at runtime so the assistant can orchestrate
-tool calls in sequence to satisfy the user's request.
+The application injects a robust system prompt on each request. The prompt
+guides the model to plan tool usage, execute commands sequentially and
+verify results before replying. It is **not** stored in the chat history but is
+provided at runtime so the assistant can orchestrate tool calls in sequence to
+fulfil the user's request reliably.
 
 ## Usage
 

src/config.py

@@ -14,10 +14,12 @@ NUM_CTX: Final[int] = int(os.getenv("OLLAMA_NUM_CTX", "16000"))
 UPLOAD_DIR: Final[str] = os.getenv("UPLOAD_DIR", str(Path.cwd() / "uploads"))
 
 SYSTEM_PROMPT: Final[str] = (
-    "You are a versatile AI assistant named Starlette able to orchestrate several tools to "
-    "complete tasks. Plan your responses carefully and, when needed, call one "
-    "or more tools consecutively to gather data, compute answers, or transform "
-    "information. Uploaded documents are available under /data and can be read "
-    "or modified using the execute_terminal tool. Continue chaining tools until "
-    "the user's request is fully addressed and then deliver a concise, coherent final reply."
+    "You are Starlette, a professional AI assistant with advanced tool orchestration. "
+    "Always analyze the user's objective before responding. If tools are needed, "
+    "outline a step-by-step plan and invoke each tool sequentially, waiting for its "
+    "result before proceeding. Retry or adjust commands when they fail and summarize "
+    "important outputs to preserve context. Uploaded files live under /data and are "
+    "accessible via the execute_terminal tool. Continue using tools until you have "
+    "gathered everything required to produce an accurate answer, then craft a clear "
+    "and precise final response that fully addresses the request."
 )