fixes

agent.py

@@ -7,6 +7,7 @@
 import globals
 import yaml
 import atexit
+import os
 
 # Load the configuration
 with open('config.yml', 'r') as f:
@@ -17,7 +18,6 @@
 pipeline_name = config['pipeline'] 
 pipeline_path = "pipelines/" + pipeline_name + ".yml"
 
-
 if wandb_enabled: # Initialize wandb if it's enabled
     wandb.init(project="OrchestrAI")
     wandb.config.wandb_enabled = wandb_enabled

ai.py

@@ -1,6 +1,7 @@
 import openai
 import yaml
 
+
 # Load the configuration
 with open('config.yml', 'r') as f:
     config = yaml.safe_load(f)
@@ -14,6 +15,7 @@
 default_frequency_penalty = config['default_frequency_penalty']
 default_presence_penalty = config['default_presence_penalty']
 
+
 # Check if the API key works
 try:
     openai.Model.list()
@@ -70,6 +72,7 @@ def generate_response(self, prompt):
             print()
 
             response_text = "".join(chat)
+
             llm_end_time_ms = round(datetime.datetime.now().timestamp() * 1000)  # logged in milliseconds
             status_code="success"
             status_message=None,
@@ -86,6 +89,7 @@ def generate_response(self, prompt):
         self.messages.append({"role": "assistant", "content": response_text})
 
         return {
+        "module_name": self.module_name,
         "response_text": response_text,
         "messages": self.messages,
         "llm_end_time_ms": llm_end_time_ms,

config.yml

@@ -1,18 +1,18 @@
-### OpenAI Key
-openai_api_key: 'YOUR_API_KEY_HERE'
+### Pipeline ###
+pipeline: 'engineering_pipeline'
+
+### API Keys ###
+openai_api_key: 'OPENAI_API_KEY'
 
 ### OpenAI Default Configs ###
-### All calls to OpenAI will use these configs unless otherwise specified ###
+### All calls to OpenAI will use these configs unless otherwise specified in pipeline ###
 
 default_model: 'gpt-4' # Recommended to set to a cheaper model for testing.
-default_temperature: 1
+default_temperature: 0.8
 default_max_tokens: Null
 default_top_p: 1
 default_frequency_penalty: 0
 default_presence_penalty: 0
 
 ### Logging ###
-wandb_enabled: false
-
-### Pipeline ###
-pipeline: 'task_pipeline'
+wandb_enabled: false

helpers.py

@@ -6,6 +6,10 @@
 import matplotlib.pyplot as plt
 import shutil
 
+#open config file
+with open('config.yml', 'r') as f:
+    config = yaml.safe_load(f)
+
 def load_pipeline(file_path):
     """Load a pipeline configuration from a YAML file."""
     print("\033[93mLoading pipeline...\033[00m")
@@ -14,15 +18,15 @@ def load_pipeline(file_path):
     return pipeline
 
 def load_system_prompt(module_name):
-
     # Load the generic system prompt
     with open('general_system.txt', 'r') as file:
         system_prompt = file.read().replace('\n', '')
 
     with open(f'system_prompts/{module_name}.txt', 'r') as file:
         module_prompt = file.read().replace('\n', '')
 
-    system_prompt = system_prompt + '\n\n --- ' + module_name.upper() + ' ---\n\n' +  module_prompt + '\n'
+    system_prompt += '\n\n --- ' + module_name.upper() + ' ---\n\n' +  module_prompt + '\n'
+
     return system_prompt
 
 def parse_chat(chat):
@@ -45,11 +49,15 @@ def to_files(files):
     if not os.path.exists("generated_code"):
         os.mkdir("generated_code")
 
-    # Create an empty requirements.txt file by default
-    with open(os.path.join("generated_code", "requirements.txt"), "w") as req_file:
-        req_file.write("")
+    # Create an empty requirements.txt file only if it doesn't exist
+    if not os.path.exists(os.path.join("generated_code", "requirements.txt")):
+        with open(os.path.join("generated_code", "requirements.txt"), "w") as req_file:
+            req_file.write("")
 
     for file_name, file_content in files:
+        # Create directories in the file path if they don't exist
+        os.makedirs(os.path.dirname(os.path.join("generated_code", file_name)), exist_ok=True)
+
         with open(os.path.join("generated_code", file_name), "w") as file:
             file.write(file_content)
 
@@ -93,27 +101,34 @@ def run_main():
 
     return process.returncode, stderr
 
-def extract_codebase(directory='generated_code', ignore_list=[]):
-    """Extracts the exsiting codebase from the generated_code directory into a condensed string."""
+
+def extract_codebase(directory='generated_code'):
+    """Extracts the existing codebase from the generated_code directory into a condensed string."""
 
+    ignore_list = ['filename_to_ignore.py', 'another_file_to_ignore.yml']
+    ignore_endings = ['.pycache', '.pyc']
+
     result_content = []
 
-    for filename in os.listdir(directory):
-        # Skip files in ignore list
-        if filename in ignore_list:
-            continue
-
-        # Only process .py, .yml, or .md files
-        if filename.endswith(('.yml', '.py', '.md')):
-            filepath = os.path.join(directory, filename)
-            try:
-                with open(filepath, 'r') as infile:
-                    # Read the contents of the file, remove line breaks and leading spaces
-                    content = infile.read().replace('\n', '').replace('\r', '')
-                    content = ' '.join(content.split())
-                    result_content.append(f"--- File Name: {filename} ---\n{content}")
-            except Exception:
-                pass
+    for root, dirs, files in os.walk(directory):
+        for filename in files:
+            # Skip files in ignore list
+            if filename in ignore_list:
+                continue
+
+            # Only process files not in ignore_endings list
+            if not any(filename.endswith(ending) for ending in ignore_endings):
+                filepath = os.path.join(root, filename)
+                try:
+                    with open(filepath, 'r') as infile:
+                        # Read the contents of the file, remove line breaks and leading spaces
+                        content = infile.read().replace('\n', '').replace('\r', '')
+                        content = ' '.join(content.split())
+                        # Get the relative path of the file
+                        relative_filepath = os.path.relpath(filepath, directory)
+                        result_content.append(f"--- File Name: {relative_filepath} ---\n{content}")
+                except Exception:
+                    pass
 
     return "\n".join(result_content)
 
@@ -139,6 +154,7 @@ def visualize_pipeline(nx, G):
 
     plt.savefig("pipeline.png", dpi=300, bbox_inches='tight')
 
+
 def list_dependencies():
     with open('generated_code/requirements.txt', 'r') as f:
         dependencies = f.read().splitlines()

modules.py

@@ -10,9 +10,7 @@
 with open('config.yml', 'r') as f:
     config = yaml.safe_load(f)
 
-# Set the wandb_enabled flag
-wandb_enabled = config['wandb_enabled']
-
+wandb_enabled = config['wandb_enabled'] # Set the wandb_enabled flag
 
 def start_module(module_input, dummy=None):   
     """This function is used to invoke the start module, to accept user input into the pipeline""" 
@@ -23,7 +21,7 @@ def start_module(module_input, dummy=None):
     output = module_input + start
 
     if wandb_enabled:
-        wb.log_tool(tool_name = module_name,
+        wb.wandb_log_tool(tool_name = module_name,
                     inputs    = {},
                     outputs   = {"original_input": output},
                     parent    = globals.chain_span,
@@ -39,7 +37,7 @@ def human_intervention(module_input, dummy=None):
     output = module_input + additional_info
 
     if wandb_enabled:
-        wb.log_tool(tool_name = module_name,
+        wb.wandb_log_tool(tool_name = module_name,
                     inputs    = {},
                     outputs   = {"human_intervention": output},
                     parent    = globals.chain_span,
@@ -60,7 +58,7 @@ def chameleon(prompt, module_name, model_config=None):
     response = ai.generate_response(prompt)
 
     if wandb_enabled:
-        wb.log_llm(response, module_name, ai.model, ai.temperature, parent = globals.chain_span)
+        wb.wandb_log_llm(response, ai.model, ai.temperature, parent = globals.chain_span)
 
     return response["response_text"]
 
@@ -76,7 +74,7 @@ def engineer(prompt, model_config=None):
     response = ai.generate_response(prompt)
 
     if wandb_enabled:
-        wb.log_llm(response, module_name, ai.model, ai.temperature, parent = globals.chain_span)
+        wb.wandb_log_llm(response, ai.model, ai.temperature, parent = globals.chain_span)
 
     response_text = response["response_text"]
 
@@ -144,7 +142,7 @@ def debugger(codebase, model_config=None):
                 debug_response = ai.generate_response(prompt)
 
                 if wandb_enabled:
-                    wb.log_llm(debug_response, module_name, ai.model, ai.temperature, parent = globals.chain_span)
+                    wb.wandb_log_llm(debug_response, ai.model, ai.temperature, parent = globals.chain_span)
 
                 debugged_code = h.parse_chat(debug_response["response_text"])
 
@@ -188,7 +186,7 @@ def modify_codebase(codebase, model_config=None):
         response = ai.generate_response(codebase)
 
         if wandb_enabled:
-            wb.log_llm(response, module_name, ai.model, ai.temperature, parent = globals.chain_span)
+            wb.wandb_log_llm(response, ai.model, ai.temperature, parent = globals.chain_span)
 
         # Parse the chat and extract files
         files = h.parse_chat(response["response_text"])
@@ -212,7 +210,7 @@ def create_readme(codebase, model_config=None):
     response = ai.generate_response(codebase)
 
     if wandb_enabled:
-        wb.log_llm(response, module_name, ai.model, ai.temperature, parent = globals.chain_span)
+        wb.wandb_log_llm(response, ai.model, ai.temperature, parent = globals.chain_span)
 
     # Save the response to a README.md file in the generated_code folder
     h.to_files([("README.md", response["response_text"])])

pipelines/engineering_pipeline.yml

@@ -4,7 +4,6 @@ pipeline:
     output_name: request
   - module: code_planner
     inputs: [request]
-    supplement: "Use only python."
     output_name: code_plan
   - module: engineer
     inputs: [request, code_plan]

plan.txt

@@ -4,20 +4,24 @@ CODE
 
 # Wandb logging
 
-- Add wandb logging to the pipeline. DONE
-- Fix the fact that the table logging breaks the pipeline logging.  DONE 
-- Generate pipeline visualisation which I can log to wandb. DONE 
+- Add wandb logging to the pipeline.                                                    DONE
+- Fix the fact that the table logging breaks the pipeline logging.                      DONE 
+- Generate pipeline visualisation which I can log to wandb.                             DONE 
 
 # General cleanup
 
+- Add the ability to run folders and create folders                                     DONE 
+- Each generated code gets it own folder
+- Regular logging
+- If debugging message is a missing file, it want juman intervention to say "I've added the file"
 - Fix the debugging so human input actually works. 
 - Add more comments and docstrings, make sure everything is clear.
 - Example pipelines  -- almost there
-- Fix realist issues. DONE 
-- Issues with structuring of system prompt. Additional user input issues. DONE 
+- Fix realist issues.                                                                   DONE 
+- Issues with structuring of system prompt. Additional user input issues.               DONE 
 - Each module should have the the original query specified separately in system prompt. DONE
-- Control temperature from pipeline file (only if specified). DONE
-- Control model used from pipeline file (only if specified, else use default gpt-4). DONE
+- Control temperature from pipeline file (only if specified).                           DONE
+- Control model used from pipeline file (only if specified, else use default gpt-4).    DONE
 
 # Documentation
 

system_prompts/debugger.txt

@@ -5,7 +5,7 @@ Your job is to fix the error, returning the full unredacted code of any scripts
 If the error is an import error, you can update requirements.txt to include the missing package.
 
 Each modified file must strictly follow a markdown code block format, where the following tokens must be replaced such that
-FILENAME is the lowercase file name including the file extension, excluding the path,
+FILENAME is the lowercase file name including the file extension, which can also include a folder it is saved within i.e test/filename.py,,
 LANG is the markup code block language for the code's language, and CODE is the code:
 
 FILENAME

system_prompts/decision_maker.txt

@@ -0,0 +1,4 @@
+You are great at making decisions.
+You are presented with evidence and discussion, and it is your job to come to a conclusion. 
+You are great at this.
+Your decisions are unbiased and fair.

system_prompts/engineer.txt

@@ -11,7 +11,7 @@ Running the full code should always be possible by running main.py. Nothing shou
 
 Then you will output the content of each file including ALL code.
 Each file must strictly follow a markdown code block format, where the following tokens must be replaced such that
-FILENAME is the lowercase file name including the file extension, excluding the path,
+FILENAME is the lowercase file name including the file extension, which can also include a folder it is saved within i.e test/filename.py,
 LANG is the markup code block language for the code's language, and CODE is the code:
 
 FILENAME

system_prompts/modify_codebase.txt

@@ -4,7 +4,7 @@ Your job is to address the feedback from the user, fixing any errors or modifyin
 You are able to create new files and modify existing files.
 
 Each new/modified file must strictly follow a markdown code block format, where the following tokens must be replaced such that
-FILENAME is the lowercase file name including the file extension, excluding the path,
+FILENAME is the lowercase file name including the file extension, which can also include a folder it is saved within i.e test/filename.py,,
 LANG is the markup code block language for the code's language, and CODE is the code:
 
 FILENAME

system_prompts/realist.txt

@@ -1,2 +1,2 @@
-You provide feedback about information that ground it in the real word and reality.
-Be harsh and consistent. 
+You provide feedback about information that grounds it in the real word and reality.
+Be harsh and consistent.

wandb_logging.py

@@ -2,21 +2,21 @@
 from wandb.sdk.data_types.trace_tree import Trace
 import time
 
-def log_llm(data, module_name, model, temperature, parent):
+def wandb_log_llm(data, model, temperature, parent):
 
     runtime = data["llm_end_time_ms"] - data["llm_start_time_ms"]
 
 
     llm_span = Trace(
-        name=module_name,
+        name=data["module_name"],
         kind="llm",
         status_code=data["status_code"],
         status_message=data["status_message"],
         metadata={
             "temperature": temperature,
             "token_usage": {"total_tokens": data["token_count"]},
             "runtime_ms": runtime,
-            "module_name": module_name,
+            "module_name": data["module_name"],
             "model_name": model
         },
         start_time_ms=parent._span.end_time_ms,
@@ -36,7 +36,7 @@ def log_llm(data, module_name, model, temperature, parent):
     llm_span.log(name="pipeline_trace")
 
 
-def log_tool(tool_name, inputs, outputs, parent, status="success"):
+def wandb_log_tool(tool_name, inputs, outputs, parent, status="success"):
     start_time_ms = round(time.time() * 1000)
     time.sleep(1)  # simulate tool execution time
     end_time_ms = round(time.time() * 1000)