🚀 Here's the PR! #22

See Sweep's progress at the progress dashboard!

⚡ Sweep Basic Tier: I'm using GPT-4. You have 5 GPT-4 tickets left for the month and 3 for the day. (tracking ID: 143dff3bc8)

For more GPT-4 tickets, visit our payment portal. For a one week free trial, try Sweep Pro (unlimited GPT-4 tickets).

Install Sweep Configs: Pull Request

[!TIP] I can email you next time I complete a pull request if you set up your email here!

Actions (click)

[ ] ↻ Restart Sweep

GitHub Actions✓

Here are the GitHub Actions logs prior to making any changes:

Sandbox logs for d0280cf

Checking autocoder/ai.py for syntax errors... ✅ autocoder/ai.py has no syntax errors! 1/1 ✓
Checking autocoder/ai.py for syntax errors...
✅ autocoder/ai.py has no syntax errors!

Sandbox passed on the latest main, so sandbox checks will be enabled for this issue.

Step 1: 🔎 Searching

I found the following snippets in your repository. I will now analyze these snippets and come up with a plan.

Some code snippets I think are relevant in decreasing order of relevance (click to expand). If some file is missing from here, you can mention the path in the ticket description.

https://github.com/artmoskvin/autocoder/blob/d0280cfec5ad26c72ccbcac20985fdcea66474af/autocoder/ai.py#L15-L21 https://github.com/artmoskvin/autocoder/blob/d0280cfec5ad26c72ccbcac20985fdcea66474af/autocoder/agent/plan.py#L35-L72

Step 2: ⌨️ Coding

[X] Modify autocoder/ai.py ✓ https://github.com/artmoskvin/autocoder/commit/3cefbbefdbf420c54d380fb23cc796886970c00e Edit
Modify autocoder/ai.py with contents:
• Add a new method `stream_call` to the `AI` class. This method will be responsible for handling streaming completions. It should accept an initial part of the messages and yield results incrementally as more messages are received or as the AI generates responses.
• The method signature should be: `def stream_call(self, initial_messages: List[BaseMessage]) -> Generator[str, List[BaseMessage], None]:`
• Inside `stream_call`, implement logic to handle initial messages and yield partial completions. The method should be designed to accept additional messages to continue generating further completions as needed.
• This change allows the AI to support streaming by processing and responding to messages in a more dynamic and interactive manner.

--- 
+++ 
@@ -20,3 +20,17 @@
     def call(self, messages: List[BaseMessage]) -> str:
         print_system_msg(f"Calling AI with prompt:\n{pprint_messages(messages)}")
         return self.model(messages).content
+    def stream_call(self, initial_messages: List[BaseMessage]) -> Generator[str, List[BaseMessage], None]:
+        print_system_msg(f"Starting streaming AI with initial prompt:\n{pprint_messages(initial_messages)}")
+        partial_completion = self.model(initial_messages).content
+        yield partial_completion
+        while True:
+            try:
+                new_messages = yield
+                if new_messages:
+                    print_system_msg(f"Received new messages for streaming AI:\n{pprint_messages(new_messages)}")
+                    partial_completion = self.model(new_messages).content
+                    yield partial_completion
+            except GeneratorExit:
+                print_system_msg("Streaming AI call ended.")
+                break

[X] Running GitHub Actions for autocoder/ai.py ✓ Edit
Check autocoder/ai.py with contents:

Ran GitHub Actions for 3cefbbefdbf420c54d380fb23cc796886970c00e:

[X] Modify autocoder/agent/plan.py ✓ https://github.com/artmoskvin/autocoder/commit/a0c3ffddd52c7cfb90111567beb98cd6bcb96cd1 Edit
Modify autocoder/agent/plan.py with contents:
• Modify the `generate_plan` and `generate_questions` methods to utilize the new `stream_call` method from the `AI` class for a streaming approach to generating plans and asking questions.
• In `generate_questions`, replace the call to `self.ai.call(self.chat)` with a loop that iterates over `self.ai.stream_call(self.chat)` to process questions in a streaming fashion.
• Similarly, in `generate_plan_from_chat`, replace the call to `self.ai.call(self.chat)` with a loop that iterates over `self.ai.stream_call(self.chat)` for generating the plan.
• These modifications will enable the `Plan` class to interact with the AI in a streaming manner, improving responsiveness and user experience by providing incremental updates and feedback as the AI processes the chat.

--- 
+++ 
@@ -1,5 +1,5 @@
 import ast
-from typing import List
+from typing import List, Generator

 from langchain.schema import SystemMessage, HumanMessage, AIMessage

@@ -62,13 +62,18 @@
     def generate_questions(self) -> List[str]:
         self.chat.append(SystemMessage(content=QUESTIONS_PROMPT))
         print_autocoder_msg("Thinking... :thinking_face:")
-        questions_str = self.ai.call(self.chat)
+        questions_str = "[]"
+        for partial_questions_str in self.ai.stream_call(self.chat):
+            questions_str = partial_questions_str
         return ast.literal_eval(questions_str)

     def generate_plan_from_chat(self) -> str:
         self.chat.append(SystemMessage(content=PLAN_PROMPT))
         print_autocoder_msg("Thinking... :thinking_face:")
-        return self.ai.call(self.chat)
+        plan = ""
+        for partial_plan in self.ai.stream_call(self.chat):
+            plan += partial_plan
+        return plan

     def ask_questions(self, questions: List[str]) -> None:
         while questions:

[X] Running GitHub Actions for autocoder/agent/plan.py ✓ Edit
Check autocoder/agent/plan.py with contents:

Ran GitHub Actions for a0c3ffddd52c7cfb90111567beb98cd6bcb96cd1:

Step 3: 🔁 Code Review

I have finished reviewing the code for completeness. I did not find errors for sweep/enable_chat_completion_streaming.

🎉 Latest improvements to Sweep:

New dashboard launched for real-time tracking of Sweep issues, covering all stages from search to coding.
Integration of OpenAI's latest Assistant API for more efficient and reliable code planning and editing, improving speed by 3x.
Use the GitHub issues extension for creating Sweep issues directly from your editor.

💡 To recreate the pull request edit the issue title or description. To tweak the pull request, leave a comment on the pull request.^{Something wrong? Let us know.}

This is an automated message generated by Sweep AI.

artmoskvin / autocoder