Adding parameter support for Local LLMs and GROQ

.gitignore

@@ -171,4 +171,5 @@ workspace
 pilot-env/
 
 # Other
-brija.py
+brija.py
+/pilot/.vs

pilot/.env.example

@@ -1,4 +1,4 @@
-# OPENAI or AZURE or OPENROUTER (ignored for Anthropic)
+# OPENAI or AZURE or OPENROUTER (ignored for Anthropic) or GROQ
 ENDPOINT=OPENAI
 
 # OPENAI_ENDPOINT=https://api.openai.com/v1/chat/completions
@@ -8,6 +8,9 @@ OPENAI_API_KEY=
 AZURE_API_KEY=
 AZURE_ENDPOINT=
 
+GROQ_API_KEY=
+GROQ_ENDPOINT=
+
 OPENROUTER_API_KEY=
 
 # Set this to use Anthropic API directly

pilot/.env.groq.example

@@ -0,0 +1,41 @@
+
+ENDPOINT=GROQ
+GROQ_API_KEY=sk-your-api-key-here
+
+# In case of Azure/OpenRouter endpoint, change this to your deployed model name
+MODEL_NAME=groq/mixtral-8x7b-32768
+# In case of Anthropic, use "anthropic/" + the model name, example for Claude 3 Opus
+# MODEL_NAME=anthropic/claude-3-opus-20240229
+MAX_TOKENS=8192
+
+# Parameters for fine tuning GROQ
+#N=2
+TEMPERATURE=0.8
+
+#TOP_P=0.95
+#FREQUENCY_PENALTY=0
+#PRESENCE_PENALTY=0
+
+API_CONNECT_TIMEOUT=30
+API_READ_TIMEOUT=3000
+
+# Folders which shouldn't be tracked in workspace (useful to ignore folders created by compiler)
+# IGNORE_PATHS=folder1,folder2
+
+# Database
+# DATABASE_TYPE=postgres
+
+DB_NAME=gpt-pilot
+DB_HOST=
+DB_PORT=
+DB_USER=
+DB_PASSWORD=
+
+OPENAI_API_KEY=sk_needs_dummy_here
+# USE_GPTPILOT_FOLDER=true
+
+# Load database imported from another location/system - EXPERIMENTAL
+# AUTOFIX_FILE_PATHS=false
+
+# Set extra buffer to wait on top of detected retry time when rate limmit is hit. defaults to 6
+# RATE_LIMIT_EXTRA_BUFFER=

pilot/.env.local_llm.example

@@ -0,0 +1,45 @@
+# OPENAI or AZURE or OPENROUTER (ignored for Anthropic)
+ENDPOINT=OPENAI
+
+# OPENAI_ENDPOINT=https://api.openai.com/v1/chat/completions
+OPENAI_ENDPOINT=http://127.0.0.1:8090/v1/chat/completions
+OPENAI_API_KEY=sk-blabla-blabla
+
+# In case of Azure/OpenRouter endpoint, change this to your deployed model name
+MODEL_NAME=gpt-4-turbo-preview
+# In case of Anthropic, use "anthropic/" + the model name, example for Claude 3 Opus
+# MODEL_NAME=anthropic/claude-3-opus-20240229
+MAX_TOKENS=8192
+
+# Parameters for fin tuning local LLMs
+TEMPERATURE=0.8
+TOP_P=0.95
+TOP_K=40
+REPETITION_PENALTY=1.1
+PRESENCE_PENALTY=0
+FREQUENCY_PENALTY=0
+# FOR GPTQ and ExLlama models
+GUIDANCE_SCALE=1.8
+
+API_CONNECT_TIMEOUT=30
+API_READ_TIMEOUT=3000
+
+# Folders which shouldn't be tracked in workspace (useful to ignore folders created by compiler)
+# IGNORE_PATHS=folder1,folder2
+
+# Database
+# DATABASE_TYPE=postgres
+
+DB_NAME=gpt-pilot
+DB_HOST=
+DB_PORT=
+DB_USER=
+DB_PASSWORD=
+
+# USE_GPTPILOT_FOLDER=true
+
+# Load database imported from another location/system - EXPERIMENTAL
+# AUTOFIX_FILE_PATHS=false
+
+# Set extra buffer to wait on top of detected retry time when rate limmit is hit. defaults to 6
+# RATE_LIMIT_EXTRA_BUFFER=

pilot/const/llm.py

@@ -3,5 +3,5 @@
 MIN_TOKENS_FOR_GPT_RESPONSE = 600
 MAX_QUESTIONS = 5
 END_RESPONSE = "EVERYTHING_CLEAR"
-API_CONNECT_TIMEOUT = 30 # timeout for connecting to the API and sending the request (seconds)
-API_READ_TIMEOUT = 300 # timeout for receiving the response (seconds)
+API_CONNECT_TIMEOUT = int(os.getenv('API_CONNECT_TIMEOUT', 30)) # timeout for connecting to the API and sending the request (seconds)
+API_READ_TIMEOUT = int(os.getenv('API_READ_TIMEOUT', 300)) # timeout for receiving the response (seconds)

pilot/helpers/AgentConvo.py

@@ -2,6 +2,7 @@
 import re
 import subprocess
 import uuid
+import os
 from os.path import sep
 
 from utils.style import color_yellow, color_yellow_bold, color_red_bold
@@ -25,14 +26,17 @@ class AgentConvo:
  agent: An instance of the agent participating in the conversation.
  """
 
- def __init__(self, agent, temperature: float = 0.7):
+ def __init__(self, agent, temperature: float = -1):
  # [{'role': 'system'|'user'|'assistant', 'content': ''}, ...]
  self.messages: list[dict] = []
  self.branches = {}
  self.log_to_user = True
  self.agent = agent
  self.high_level_step = self.agent.project.current_step
- self.temperature = temperature
+ if (temperature < 0.0): 
+ self.temperature = float(os.getenv('TEMPERATURE', 0.7))
+ else:
+ self.temperature = temperature
 
  # add system message
  system_message = get_sys_message(self.agent.role, self.agent.project.args)

pilot/helpers/agents/CodeMonkey.py

@@ -182,10 +182,11 @@ def remove_backticks(content: str) -> str:
  :param content: content to remove backticks from
  :return: content without backticks
  """
- start_pattern = re.compile(r"^\s*```([a-z0-9]+)?\n")
- end_pattern = re.compile(r"\n```\s*$")
- content = start_pattern.sub("", content)
- content = end_pattern.sub("", content)
+ pattern = r"^.*```([a-z0-9]+)?\n(.*)?```.*$"
+ match = re.search(pattern, content, re.DOTALL)
+ if match:
+ code = match.group(2)
+ return code.strip()
  return content
 
  def identify_file_to_change(self, code_changes_description: str, files: list[dict]) -> str:

pilot/prompts/development/parse_task.prompt

@@ -32,7 +32,7 @@ Examples:
  "type": "command",
  "command": {
  "command": "mv ./index.js ./public/index.js"",
- "timeout": "5000",
+ "timeout": 5000,
  "success_message": "",
  "command_id": "move_index_file"
  }

pilot/utils/function_calling.py

@@ -55,7 +55,9 @@ def add_function_calls_to_request(gpt_data, function_calls: Union[FunctionCallSe
  'content': prompter.prompt('', function_calls['definitions'], function_call)
  }
  gpt_data['messages'].append(function_call_message)
-
+
+ gpt_data['response_format'] = {"type": "json_object"}
+
  return function_call_message
 
 
@@ -71,7 +73,15 @@ def parse_agent_response(response, function_calls: Union[FunctionCallSet, None])
  """
  if function_calls:
  text = response['text']
- return json.loads(text)
+ try:
+ return json.loads(text)
+ except:
+ pattern = r'```(.*)\n(.*?)```'
+ match = re.search(pattern, text, re.DOTALL)
+
+ if match:
+ code = match.group(1)
+ return json.loads(code.strip())
 
  return response['text']