Merge branch 'main' into config

README.md

@@ -145,12 +145,16 @@ To start using Devika, follow these steps:
 
 Devika requires certain configuration settings and API keys to function properly. Update the `config.toml` file with the following information:
 
-- `OPENAI_API_KEY`: Your OpenAI API key for accessing GPT models.
-- `CLAUDE_API_KEY`: Your Anthropic API key for accessing Claude models.
-- `BING_API_KEY`: Your Bing Search API key for web searching capabilities.
-- `DATABASE_URL`: The URL for your database connection.
-- `LOG_DIRECTORY`: The directory where Devika's logs will be stored.
-- `PROJECT_DIRECTORY`: The directory where Devika's projects will be stored.
+- `SQLITE_DB`: The path to the SQLite database file for storing Devika's data.
+- `SCREENSHOTS_DIR`: The directory where screenshots captured by Devika will be stored.
+- `PDFS_DIR`: The directory where PDF files processed by Devika will be stored.
+- `PROJECTS_DIR`: The directory where Devika's projects will be stored.
+- `LOGS_DIR`: The directory where Devika's logs will be stored.
+- `REPOS_DIR`: The directory where Git repositories cloned by Devika will be stored.
+- `BING`: Your Bing Search API key for web searching capabilities.
+- `CLAUDE`: Your Anthropic API key for accessing Claude models.
+- `NETLIFY`: Your Netlify API key for deploying and managing web projects.
+- `OPENAI`: Your OpenAI API key for accessing GPT models.
 
 Make sure to keep your API keys secure and do not share them publicly.
 

config.toml

@@ -11,6 +11,7 @@ BING = "<YOUR_BING_API_KEY>"
 CLAUDE = "<YOUR_CLAUDE_API_KEY>"
 NETLIFY = "<YOUR_NETLIFY_API_KEY>"
 OPENAI = "<YOUR_OPENAI_API_KEY>"
+GROQ = "<YOUR_GROQ_API_KEY>"
 
 [API_ENDPOINTS]
 BING = "https://api.bing.microsoft.com/v7.0/search"

devika.py

@@ -160,8 +160,9 @@ def calculate_tokens():
 @app.route("/api/token-usage", methods=["GET"])
 @route_logger(logger)
 def token_usage():
-    from src.llm import TOKEN_USAGE
-    return jsonify({"token_usage": TOKEN_USAGE})
+    project_name = request.args.get("project_name")
+    token_count = AgentState().get_latest_token_usage(project_name)
+    return jsonify({"token_usage": token_count})
 
 
 @app.route("/api/real-time-logs", methods=["GET"])

requirements.txt

@@ -20,4 +20,5 @@ keybert
 GitPython
 netlify-py
 Markdown
-xhtml2pdf
+xhtml2pdf
+groq

src/agents/action/action.py

@@ -39,15 +39,15 @@ def validate_response(self, response: str):
         else:
             return response["response"], response["action"]
 
-    def execute(self, conversation: list) -> str:
+    def execute(self, conversation: list, project_name: str) -> str:
         prompt = self.render(conversation)
-        response = self.llm.inference(prompt)
+        response = self.llm.inference(prompt, project_name)
 
         valid_response = self.validate_response(response)
 
         while not valid_response:
             print("Invalid response from the model, trying again...")
-            return self.execute(conversation)
+            return self.execute(conversation, project_name)
 
         print("===" * 10)
         print(valid_response)

src/agents/agent.py

@@ -93,7 +93,8 @@ def search_queries(self, queries: list, project_name: str) -> dict:
             Formatter Agent is invoked to format and learn from the contents
             """
             results[query] = self.formatter.execute(
-                browser.extract_text()
+                browser.extract_text(),
+                project_name
             )
 
             """
@@ -118,7 +119,7 @@ def update_contextual_keywords(self, sentence: str):
     Decision making Agent
     """
     def make_decision(self, prompt: str, project_name: str) -> str:
-        decision = self.decision.execute(prompt)
+        decision = self.decision.execute(prompt, project_name)
 
         for item in decision:
             function = item["function"]
@@ -134,7 +135,7 @@ def make_decision(self, prompt: str, project_name: str) -> str:
             elif function == "generate_pdf_document":
                 user_prompt = args["user_prompt"]
                 # Call the reporter agent to generate the PDF document
-                markdown = self.reporter.execute([user_prompt], "")
+                markdown = self.reporter.execute([user_prompt], "", project_name)
                 _out_pdf_file = PDF().markdown_to_pdf(markdown, project_name)
 
                 project_name_space_url = project_name.replace(" ", "%20")
@@ -154,10 +155,10 @@ def make_decision(self, prompt: str, project_name: str) -> str:
             elif function == "coding_project":
                 user_prompt = args["user_prompt"]
                 # Call the planner, researcher, coder agents in sequence
-                plan = self.planner.execute(user_prompt)
+                plan = self.planner.execute(user_prompt, project_name)
                 planner_response = self.planner.parse_response(plan)
 
-                research = self.researcher.execute(plan, self.collected_context_keywords)
+                research = self.researcher.execute(plan, self.collected_context_keywords, project_name)
                 search_results = self.search_queries(research["queries"], project_name)
 
                 code = self.coder.execute(
@@ -177,7 +178,7 @@ def subsequent_execute(self, prompt: str, project_name: str) -> str:
         conversation = ProjectManager().get_all_messages_formatted(project_name)
         code_markdown = ReadCode(project_name).code_set_to_markdown()
 
-        response, action = self.action.execute(conversation)
+        response, action = self.action.execute(conversation, project_name)
 
         ProjectManager().add_message_from_devika(project_name, response)
 
@@ -188,7 +189,8 @@ def subsequent_execute(self, prompt: str, project_name: str) -> str:
         if action == "answer":
             response = self.answer.execute(
                 conversation=conversation,
-                code_markdown=code_markdown
+                code_markdown=code_markdown,
+                project_name=project_name
             )
             ProjectManager().add_message_from_devika(project_name, response)
         elif action == "run":
@@ -238,7 +240,7 @@ def subsequent_execute(self, prompt: str, project_name: str) -> str:
 
             self.patcher.save_code_to_project(code, project_name)
         elif action == "report":
-            markdown = self.reporter.execute(conversation, code_markdown)
+            markdown = self.reporter.execute(conversation, code_markdown, project_name)
 
             _out_pdf_file = PDF().markdown_to_pdf(markdown, project_name)
 
@@ -261,7 +263,7 @@ def execute(self, prompt: str, project_name_from_user: str = None) -> str:
         if project_name_from_user:
             ProjectManager().add_message_from_user(project_name_from_user, prompt)
 
-        plan = self.planner.execute(prompt)
+        plan = self.planner.execute(prompt, project_name_from_user)
         print(plan)
         print("=====" * 10)
 
@@ -288,15 +290,15 @@ def execute(self, prompt: str, project_name_from_user: str = None) -> str:
         self.update_contextual_keywords(focus)
         print(self.collected_context_keywords)
 
-        internal_monologue = self.internal_monologue.execute(current_prompt=plan)
+        internal_monologue = self.internal_monologue.execute(current_prompt=plan, project_name=project_name)
         print(internal_monologue)
         print("=====" * 10)
 
         new_state = AgentState().new_state()
         new_state["internal_monologue"] = internal_monologue
         AgentState().add_to_current_state(project_name, new_state)
 
-        research = self.researcher.execute(plan, self.collected_context_keywords)
+        research = self.researcher.execute(plan, self.collected_context_keywords, project_name)
         print(research)
         print("=====" * 10)
 

src/agents/answer/answer.py

@@ -40,14 +40,14 @@ def validate_response(self, response: str):
         else:
             return response["response"]
 
-    def execute(self, conversation: list, code_markdown: str) -> str:
+    def execute(self, conversation: list, code_markdown: str, project_name: str) -> str:
         prompt = self.render(conversation, code_markdown)
-        response = self.llm.inference(prompt)
+        response = self.llm.inference(prompt, project_name)
 
         valid_response = self.validate_response(response)
 
         while not valid_response:
             print("Invalid response from the model, trying again...")
-            return self.execute(conversation, code_markdown)
+            return self.execute(conversation, code_markdown, project_name)
 
         return valid_response

src/agents/coder/coder.py

@@ -102,13 +102,13 @@ def execute(
         project_name: str
     ) -> str:
         prompt = self.render(step_by_step_plan, user_context, search_results)
-        response = self.llm.inference(prompt)
+        response = self.llm.inference(prompt, project_name)
 
         valid_response = self.validate_response(response)
 
         while not valid_response:
             print("Invalid response from the model, trying again...")
-            return self.execute(step_by_step_plan, user_context, search_results)
+            return self.execute(step_by_step_plan, user_context, search_results, project_name)
 
         print(valid_response)
 

src/agents/decision/decision.py

@@ -32,14 +32,14 @@ def validate_response(self, response: str):
 
         return response
 
-    def execute(self, prompt: str) -> str:
+    def execute(self, prompt: str, project_name: str) -> str:
         prompt = self.render(prompt)
-        response = self.llm.inference(prompt)
+        response = self.llm.inference(prompt, project_name)
 
         valid_response = self.validate_response(response)
 
         while not valid_response:
             print("Invalid response from the model, trying again...")
-            return self.execute(prompt)
+            return self.execute(prompt, project_name)
 
         return valid_response

src/agents/feature/feature.py

@@ -103,7 +103,7 @@ def execute(
         project_name: str
     ) -> str:
         prompt = self.render(conversation, code_markdown, system_os)
-        response = self.llm.inference(prompt)
+        response = self.llm.inference(prompt, project_name)
 
         valid_response = self.validate_response(response)
 

src/agents/formatter/formatter.py

@@ -16,7 +16,7 @@ def render(self, raw_text: str) -> str:
     def validate_response(self, response: str) -> bool:
         return True
 
-    def execute(self, raw_text: str) -> str:
+    def execute(self, raw_text: str, project_name: str) -> str:
         raw_text = self.render(raw_text)
-        response = self.llm.inference(raw_text)
+        response = self.llm.inference(raw_text, project_name)
         return response

src/agents/internal_monologue/internal_monologue.py

@@ -33,15 +33,15 @@ def validate_response(self, response: str):
         else:
             return response["internal_monologue"]
 
-    def execute(self, current_prompt: str) -> str:
+    def execute(self, current_prompt: str, project_name: str) -> str:
         current_prompt = self.render(current_prompt)
-        response = self.llm.inference(current_prompt)
+        response = self.llm.inference(current_prompt, project_name)
 
         valid_response = self.validate_response(response)
 
         while not valid_response:
             print("Invalid response from the model, trying again...")
-            return self.execute(current_prompt)
+            return self.execute(current_prompt, project_name)
 
         return valid_response
 

src/agents/patcher/patcher.py

@@ -115,7 +115,7 @@ def execute(
             error,
             system_os
         )
-        response = self.llm.inference(prompt)
+        response = self.llm.inference(prompt, project_name)
 
         valid_response = self.validate_response(response)
 

src/agents/planner/planner.py

@@ -65,7 +65,7 @@ def parse_response(self, response: str):
 
         return result    
 
-    def execute(self, prompt: str) -> str:
+    def execute(self, prompt: str, project_name: str) -> str:
         prompt = self.render(prompt)
-        response = self.llm.inference(prompt)
+        response = self.llm.inference(prompt, project_name)
         return response

src/agents/reporter/reporter.py

@@ -28,16 +28,17 @@ def validate_response(self, response: str):
 
     def execute(self,
         conversation: list,
-        code_markdown: str
+        code_markdown: str,
+        project_name: str
     ) -> str:
         prompt = self.render(conversation, code_markdown)
-        response = self.llm.inference(prompt)
+        response = self.llm.inference(prompt, project_name)
 
         valid_response = self.validate_response(response)
 
         while not valid_response:
             print("Invalid response from the model, trying again...")
-            return self.execute(conversation, code_markdown)
+            return self.execute(conversation, code_markdown, project_name)
 
         return valid_response
 

src/agents/researcher/researcher.py

@@ -42,16 +42,16 @@ def validate_response(self, response: str):
                 "ask_user": response["ask_user"]
             }
 
-    def execute(self, step_by_step_plan: str, contextual_keywords: List[str]) -> str:
+    def execute(self, step_by_step_plan: str, contextual_keywords: List[str], project_name: str) -> str:
         contextual_keywords = ", ".join(map(lambda k: k.capitalize(), contextual_keywords))
         step_by_step_plan = self.render(step_by_step_plan, contextual_keywords)
 
-        response = self.llm.inference(step_by_step_plan)
+        response = self.llm.inference(step_by_step_plan, project_name)
 
         valid_response = self.validate_response(response)
 
         while not valid_response:
             print("Invalid response from the model, trying again...")
-            return self.execute(step_by_step_plan, contextual_keywords)
+            return self.execute(step_by_step_plan, contextual_keywords, project_name)
 
         return valid_response

src/agents/runner/runner.py

@@ -136,7 +136,7 @@ def run_code(
                     error=command_output
                 )
 
-                response = self.llm.inference(prompt)
+                response = self.llm.inference(prompt, project_name)
 
                 valid_response = self.validate_rerunner_response(response)
 
@@ -233,7 +233,7 @@ def execute(
         project_name: str
     ) -> str:
         prompt = self.render(conversation, code_markdown, os_system)
-        response = self.llm.inference(prompt)
+        response = self.llm.inference(prompt, project_name)
 
         valid_response = self.validate_response(response)
 

src/config.py

@@ -3,6 +3,15 @@
 
 
 class Config:
+
+    _instance = None
+
+    def __new__(cls):
+        if cls._instance is None:
+            cls._instance = super().__new__(cls)
+            cls._instance.config = toml.load("config.toml")
+        return cls._instance
+
     def __init__(self):
         self.config = toml.load("config.toml")
 
@@ -28,7 +37,10 @@ def get_openai_api_key(self):
 
     def get_netlify_api_key(self):
         return environ.get("NETLIFY_API_KEY", self.config["API_KEYS"]["NETLIFY"])
-
+
+    def get_groq_api_key(self):
+        return environ.get("GROQ_API_KEY", self.config["API_KEYS"]["GROQ"])
+
     def get_sqlite_db(self):
         return environ.get("SQLITE_DB_PATH", self.config["STORAGE"]["SQLITE_DB"])
 

src/llm/__init__.py

@@ -1 +1 @@
-from .llm import LLM, TOKEN_USAGE
+from .llm import LLM

src/llm/groq_client.py

@@ -0,0 +1,24 @@
+from groq import Groq
+
+from src.config import Config
+
+class Groq:
+    def __init__(self, api_key: str):
+        config = Config()
+        api_key = config.get_groq_api_key()
+        self.client = Groq(
+            api_key=api_key,
+        )
+
+    def inference(self, model_id: str, prompt: str) -> str:
+        chat_completion = self.client.chat.completions.create(
+            messages=[
+                {
+                    "role": "user",
+                    "content": prompt.strip(),
+                }
+            ],
+            model=model_id,
+        )
+
+        return chat_completion.choices[0].message.content