wip

.github/workflows/main-release.yaml

@@ -8,6 +8,7 @@ on:
   push:
     branches:
       - main
+      - fork-modifications
 
 jobs:
   build-and-push:
@@ -32,13 +33,22 @@ jobs:
       - name: Expose GH Runtime
         uses: crazy-max/ghaction-github-runtime@v3
 
+      - name: set lower case owner name
+        run: |
+          echo "OWNER_LC=${OWNER,,}" >>${GITHUB_ENV}
+        env:
+          OWNER: "${{ github.repository_owner }}"
+
+      - name: Convert repository name to lowercase
+        run: echo "REPO_NAME_LC=$(echo ${{ github.event.repository.name }} | awk '{print tolower($0)}')" >> $GITHUB_ENV
+
       - name: Build and Push Docker Images
         run: |
           make build_and_push_images
         env:
           REGISTRY: "ghcr.io"
-          ORG: ${{ github.repository_owner }}
-          REPO: ${{ github.event.repository.name }}
+          ORG: ${{ env.OWNER_LC }}
+          REPO: ${{ env.REPO_NAME_LC }}
           GITHUB_WORKFLOW: ${{ github.workflow }}
 
   build-tauri:

.github/workflows/tag-release.yaml

@@ -30,7 +30,7 @@ jobs:
 
       - name: Build and Push Docker Images
         run: |
-          TAG=${GITHUB_REF#refs/tags/} make build_and_push_images
+          TAG=${GITHUB_REF#refs/heads/} make build_and_push_images
         env:
           REGISTRY: ghcr.io
           ORG: ${{ github.repository_owner }}
@@ -69,7 +69,7 @@ jobs:
 
       - name: get release version
         id: get_release_version
-        run: echo "TAG=${GITHUB_REF#refs/tags/}" >>  $GITHUB_ENV
+        run: echo "TAG=${GITHUB_REF#refs/heads/}" >>  $GITHUB_ENV
 
       - name: get release id
         id: get_release_id

Makefile

@@ -3,7 +3,7 @@
 TAG := $(or $(TAG),main)
 GITHUB_WORKFLOW := $(or $(GITHUB_WORKFLOW),local)
 REGISTRY := $(or $(REGISTRY),index.docker.io)
-PLATFORMS := linux/amd64,linux/arm64
+PLATFORMS := linux/amd64
 BUILDX_FLAGS := --platform $(PLATFORMS) --push
 
 define get_full_tag

core/config/__init__.py

@@ -0,0 +1 @@
+from .config import *

core/config/config.py

@@ -0,0 +1,84 @@
+from os import getenv
+
+def get_mongo_database_name():
+    return getenv("MONGODB_DATABASE", "rubra_db")
+
+def get_mongo_url() -> str:
+    url = getenv("MONGODB_URL")
+    if url:
+        return url
+
+    host = getenv("MONGODB_HOST", "localhost")
+    user = getenv("MONGODB_USER", getenv("MONGODB_USERNAME", None))
+    password = getenv("MONGODB_PASS", getenv("MONGODB_PASSWORD", None))
+    port = getenv("MONGODB_PORT", 27017)
+    database = get_mongo_database_name()
+
+    if user and not password:
+        print("MONGODB_USER set but password not found, ignoring user")
+
+    if not user and password:
+        print("MONGODB_PASSWORD set but user not found, ignoring password")
+
+    if user and password:
+        return f"mongodb://{user}:{password}@{host}:${port}/{database}"
+
+    return f"mongodb://{host}:{port}/{database}"
+
+def get_redis_url() -> str:
+    url = getenv("REDIS_URL")
+    if url:
+        return url
+
+    host = getenv("REDIS_HOST", "localhost")
+    password = getenv("REDIS_PASS", getenv("REDIS_PASSWORD", None))
+    user = getenv("REDIS_USER", getenv("REDIS_USERNAME", None))
+    port = getenv("REDIS_PORT", 6379)
+    database = getenv("REDIS_DATABASE", 0)
+
+    if password:
+        return f"redis://{user or ''}:{password}@{host}:{port}/{database}"
+
+    return f"redis://{host}:{port}/{database}" 
+
+def get_litellm_url() -> str:
+    url = getenv("LITELLM_URL")
+    if url:
+        return url
+
+    host = getenv("LITELLM_HOST", "localhost")
+    port = getenv("LITELLM_PORT", 8002)
+
+    return f"http://{host}:{port}"
+
+def get_vector_db_url() -> str:
+    url = getenv("VECTOR_DB_URL")
+    if url:
+        return url
+
+    host = getenv("VECTOR_DB_HOST", "localhost")
+    port = getenv("VECTOR_DB_PORT", 8010)
+
+    return f"http://{host}:{port}"
+
+def get_embedding_url():
+    url = getenv("EMBEDDING_URL")
+    if url:
+        return url
+
+    host = getenv("EMBEDDING_HOST", "localhost")
+    port = getenv("EMBEDDING_PORT", 8020)
+
+    return f"http://{host}:{port}"
+
+mongo_database = get_mongo_database_name()
+
+mongo_url = get_mongo_url()
+
+litellm_url = get_litellm_url()
+
+vector_db_url = get_vector_db_url()
+
+redis_url = get_redis_url()
+
+embedding_url = get_embedding_url()

core/local_model.py

@@ -17,6 +17,8 @@
 )
 from openai import OpenAI
 
+import core.config as configs
+
 ner = spacy.load("en_core_web_sm")
 
 pattern = r">(.*?)</"
@@ -27,14 +29,15 @@
 TOOL_OUTPUT_ROLE = Role7.tool_output.value
 QUERY_FORMAT = '{"query": "refined question"}'
 
-litellm_host = os.getenv("LITELLM_HOST", "localhost")
-
-
 oai_client = OpenAI(
-    base_url=f"http://{litellm_host}:8002/v1/",
-    # base_url="http://localhost:1234/v1/",
-    api_key="abc",
+    base_url=f"{configs.litellm_url}/v1",
+    api_key='sk-something',
 )
+
+print(os.getenv("LITELLM_MASTER_KEY", None))
+
+print(configs.litellm_url)
+
 model_name = "custom"
 
 logging.basicConfig(
@@ -331,8 +334,17 @@ def chat(
             ):
                 messages.append(msg)
 
+        oai_client = OpenAI(
+            base_url=f"{configs.litellm_url}/v1",
+            api_key='sk-something',
+        )
+
+        print(oai_client)
+
+        os.environ["OPENAI_API_KEY"] = "sk-something"
+
         response = oai_client.chat.completions.create(
-            model="openai/custom",
+            model="ollama/llama3",
             messages=messages,
             stream=stream,
             temperature=0.1,

core/tasks/is_ready.py

@@ -0,0 +1,30 @@
+import socket
+
+import requests
+
+from core.config import litellm_url, vector_db_url
+
+def is_ready():
+    # response = requests.get(f"{litellm_url}/health", headers={
+    #     "Authorization": f"Bearer {os.getenv('LITELLM_MASTER_KEY', '')}"
+    # })
+    # if not response.ok:
+    #     raise Exception(response.text)
+
+    # print(response)
+
+    response = requests.get(f"{litellm_url}/health/readiness")
+    if not response.ok:
+        raise Exception(response.text)
+
+    print(response)
+
+    response = requests.get(f"{vector_db_url}/healthz")
+    if not response.ok:
+        raise Exception(response.text)
+
+    print(response)
+
+
+if __name__ == "__main__":
+    is_ready()

core/tasks/tasks.py

@@ -5,6 +5,16 @@
 import sys
 from functools import partial
 
+from typing import cast, Any
+
+# Third Party
+from core.tools.knowledge.vector_db.milvus.operations import add_texts, milvus_connection_alias
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from core.tools.knowledge.file_knowledge_tool import FileKnowledgeTool
+from core.tools.web_browse.web_browse_tool import WebBrowseTool
+
+from pymilvus import connections
+
 # Get the current working directory
 current_directory = os.getcwd()
 
@@ -31,33 +41,38 @@
 from openai import OpenAI
 from pymongo import MongoClient
 
-litellm_host = os.getenv("LITELLM_HOST", "localhost")
-redis_host = os.getenv("REDIS_HOST", "localhost")
-mongodb_host = os.getenv("MONGODB_HOST", "localhost")
+import core.config as configs
 
-redis_client = redis.Redis(host=redis_host, port=6379, db=0)
-app = Celery("tasks", broker=f"redis://{redis_host}:6379/0")
-app.config_from_object("core.tasks.celery_config")
-app.autodiscover_tasks(["core.tasks"])  # Explicitly discover tasks in 'app' package
+from .is_ready import is_ready
+
+redis_client = cast(redis.Redis, redis.Redis.from_url(configs.redis_url)) # annoyingly from_url returns None, not Self
+app = Celery("tasks", broker=configs.redis_url)
 
-# MongoDB Configuration
-MONGODB_URL = f"mongodb://{mongodb_host}:27017"
-DATABASE_NAME = "rubra_db"
+app.autodiscover_tasks(["core.tasks"])  # Explicitly discover tasks in 'app' package
 
 # Global MongoDB client
-mongo_client = None
+mongo_client: MongoClient = MongoClient(configs.mongo_url)
+
+def ping_pong():
+    pong = app.control.ping([f'celery@{socket.gethostname()}'])
+    if len(pong) == 0 or list(pong[0].values())[0].get('ok', None) is None:
+        raise Exception('ping failed with' + str(pong))
+
+    print(pong)
 
 
 @signals.worker_process_init.connect
-def setup_mongo_connection(*args, **kwargs):
-    global mongo_client
-    mongo_client = MongoClient(f"mongodb://{mongodb_host}:27017")
+async def ensure_connections(*args, **kwargs):
+    mongo_client.admin.command('ping')
 
+    is_ready()
+
+    ping_pong()
 
 def create_assistant_message(
     thread_id, assistant_id, run_id, content_text, role=Role7.assistant.value
 ):
-    db = mongo_client[DATABASE_NAME]
+    db = mongo_client[configs.mongo_database]
 
     # Generate a unique ID for the message
     message_id = f"msg_{uuid.uuid4().hex[:6]}"
@@ -175,10 +190,6 @@ def rubra_local_agent_chat_completion(
 
 
 def form_openai_tools(tools, assistant_id: str):
-    # Third Party
-    from core.tools.knowledge.file_knowledge_tool import FileKnowledgeTool
-    from core.tools.web_browse.web_browse_tool import WebBrowseTool
-
     retrieval = FileKnowledgeTool()
     googlesearch = WebBrowseTool()
     res_tools = []
@@ -214,13 +225,13 @@ def form_openai_tools(tools, assistant_id: str):
 
 @shared_task
 def execute_chat_completion(assistant_id, thread_id, redis_channel, run_id):
+    db = mongo_client[configs.mongo_database]
     try:
         oai_client = OpenAI(
-            base_url=f"http://{litellm_host}:8002/v1/",
-            api_key="abc",  # point to litellm server
+            base_url=configs.litellm_url,
+            api_key=os.getenv("LITELLM_MASTER_KEY"),  # point to litellm server
         )
-        db = mongo_client[DATABASE_NAME]
-
+        print(oai_client.models.list().data)
         # Fetch assistant and thread messages synchronously
         assistant = db.assistants.find_one({"id": assistant_id})
         thread_messages = list(db.messages.find({"thread_id": thread_id}))
@@ -453,15 +464,8 @@ def execute_chat_completion(assistant_id, thread_id, redis_channel, run_id):
 
 @app.task
 def execute_asst_file_create(file_id: str, assistant_id: str):
-    # Standard Library
-    import json
-
-    # Third Party
-    from core.tools.knowledge.vector_db.milvus.operations import add_texts
-    from langchain.text_splitter import RecursiveCharacterTextSplitter
-
     try:
-        db = mongo_client[DATABASE_NAME]
+        db = mongo_client[configs.mongo_database]
         collection_name = assistant_id
         text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=0)
         parsed_text = ""

core/tools/knowledge/file_knowledge_tool.py

@@ -1,13 +1,12 @@
 # Standard Library
 import json
-import os
+
+import core.config as configs
 
 # Third Party
 import requests
 
-VECTOR_DB_HOST = os.getenv("VECTOR_DB_HOST", "localhost")
-VECTOR_DB_MATCH_URL = f"http://{VECTOR_DB_HOST}:8010/similarity_match"
-
+vector_db_url = f"{configs.vector_db_url}/similarity_search"
 
 class FileKnowledgeTool:
     name = "FileKnowledge"
@@ -42,7 +41,7 @@ def file_knowledge_search_api(query: str, assistant_id: str):
         }
     )
 
-    response = requests.post(VECTOR_DB_MATCH_URL, headers=headers, data=data)
+    response = requests.post(vector_db_url, headers=headers, data=data)
     res = response.json()["response"]
     txt = ""
     for r in res:

core/tools/knowledge/vector_db/milvus/custom_embeddigs.py

@@ -1,14 +1,14 @@
 # Standard Library
 import json
-import os
 from typing import List
 
 # Third Party
 import requests
 from langchain.embeddings.base import Embeddings
 
-HOST = os.getenv("EMBEDDING_HOST", "localhost")
-EMBEDDING_URL = f"http://{HOST}:8020/embed_multiple"
+import core.config as configs
+
+EMBEDDING_URL = f"{configs.embedding_url}/embed_multiple"
 
 
 def embed_text(texts: List[str]) -> List[List[float]]: