g4f fixed

README.md

@@ -153,12 +153,12 @@ Enable or disable plugins. Default value True.
 Add your userid from telegram. If empty DM enable or disable option will be disabled.     
 
 - `GPT_KEY`
-To use GPT4free GPT_KEY = '' (Only working in docker)
+To use GPT4free GPT_KEY = ''        
 Key from the provider (including openai). Whichever api base you want use the key provided.
 For Naga AI, Get this by messaging run the /key get command in th bot channel in [Naga AI Discord](https://discord.gg/JxRBXBhabu) , [Naga AI Telegram](https://t.me/chimer_ai)     
 
 - `API_BASE`
-To use GPT4free API_BASE = 'http://g4f_server:1337' (Only working in docker)
+To use GPT4free API_BASE = 'http://localhost:1337'      
 You can use any provider. I have included Naga AI api base. Use the key for the same.
 
 - `MAX_HISTORY`

docker-compose.yml

@@ -9,15 +9,3 @@ services:
     restart: always
 #    volumes:
 #      - ./home/personas:/app/personas
-#    depends_on:
-#      - g4f_server
-#  g4f_server:
-#    container_name: g4f_server
-#    ports:
-#      - '1337:1337'
-#    environment:
-#      - PYTHONUNBUFFERED=1
-#    build:
-#      context: ./interference
-#      dockerfile: Dockerfile
-#    restart: always

interference/app.py

@@ -7,14 +7,17 @@
 from typing       import Any
 from flask        import Flask, jsonify, request
 from flask_cors   import CORS
-from transformers import AutoTokenizer
+#from transformers import AutoTokenizer
 import uvicorn
 from g4f          import ChatCompletion
 
 
 app = Flask(__name__)
 CORS(app)
 
+@app.route('/')
+def index():
+    return 'interference api, url: http://127.0.0.1:1337'
 
 @app.route('/models')
 def get_models():
@@ -119,71 +122,72 @@ def streaming():
 
 
 # Get the embedding from huggingface
-def get_embedding(input_text, token):
-    huggingface_token = token
-    embedding_model = 'sentence-transformers/all-mpnet-base-v2'
-    max_token_length = 500
-
-    # Load the tokenizer for the 'all-mpnet-base-v2' model
-    tokenizer = AutoTokenizer.from_pretrained(embedding_model)
-    # Tokenize the text and split the tokens into chunks of 500 tokens each
-    tokens = tokenizer.tokenize(input_text)
-    token_chunks = [tokens[i:i + max_token_length]
-                    for i in range(0, len(tokens), max_token_length)]
-
-    # Initialize an empty list
-    embeddings = []
-
-    # Create embeddings for each chunk
-    for chunk in token_chunks:
-        # Convert the chunk tokens back to text
-        chunk_text = tokenizer.convert_tokens_to_string(chunk)
-
-        # Use the Hugging Face API to get embeddings for the chunk
-        api_url = f'https://api-inference.huggingface.co/pipeline/feature-extraction/{embedding_model}'
-        headers = {'Authorization': f'Bearer {huggingface_token}'}
-        chunk_text = chunk_text.replace('\n', ' ')
-
-        # Make a POST request to get the chunk's embedding
-        response = requests.post(api_url, headers=headers, json={
-                                 'inputs': chunk_text, 'options': {'wait_for_model': True}})
-
-        # Parse the response and extract the embedding
-        chunk_embedding = response.json()
-        # Append the embedding to the list
-        embeddings.append(chunk_embedding)
-
-    # averaging all the embeddings
-    # this isn't very effective
-    # someone a better idea?
-    num_embeddings = len(embeddings)
-    average_embedding = [sum(x) / num_embeddings for x in zip(*embeddings)]
-    embedding = average_embedding
-    return embedding
-
-
-@app.route('/embeddings', methods=['POST'])
-def embeddings():
-    input_text_list = request.get_json().get('input')
-    input_text      = ' '.join(map(str, input_text_list))
-    token           = request.headers.get('Authorization').replace('Bearer ', '')
-    embedding       = get_embedding(input_text, token)
-
-    return {
-        'data': [
-            {
-                'embedding': embedding,
-                'index': 0,
-                'object': 'embedding'
-            }
-        ],
-        'model': 'text-embedding-ada-002',
-        'object': 'list',
-        'usage': {
-            'prompt_tokens': None,
-            'total_tokens': None
-        }
-    }
+#def get_embedding(input_text, token):
+#    huggingface_token = token
+#    embedding_model = 'sentence-transformers/all-mpnet-base-v2'
+#    max_token_length = 500
+#
+#    # Load the tokenizer for the 'all-mpnet-base-v2' model
+#    tokenizer = AutoTokenizer.from_pretrained(embedding_model)
+#    # Tokenize the text and split the tokens into chunks of 500 tokens each
+#    tokens = tokenizer.tokenize(input_text)
+#    token_chunks = [tokens[i:i + max_token_length]
+#                    for i in range(0, len(tokens), max_token_length)]
+#
+#    # Initialize an empty list
+#    embeddings = []
+#
+#    # Create embeddings for each chunk
+#    for chunk in token_chunks:
+#        # Convert the chunk tokens back to text
+#        chunk_text = tokenizer.convert_tokens_to_string(chunk)
+#
+#        # Use the Hugging Face API to get embeddings for the chunk
+#        api_url = f'https://api-inference.huggingface.co/pipeline/feature-extraction/{embedding_model}'
+#        headers = {'Authorization': f'Bearer {huggingface_token}'}
+#        chunk_text = chunk_text.replace('\n', ' ')
+#
+#        # Make a POST request to get the chunk's embedding
+#        response = requests.post(api_url, headers=headers, json={
+#                                 'inputs': chunk_text, 'options': {'wait_for_model': True}})
+#
+#        # Parse the response and extract the embedding
+#        chunk_embedding = response.json()
+#        # Append the embedding to the list
+#        embeddings.append(chunk_embedding)
+#
+#    # averaging all the embeddings
+#    # this isn't very effective
+#    # someone a better idea?
+#    num_embeddings = len(embeddings)
+#    average_embedding = [sum(x) / num_embeddings for x in zip(*embeddings)]
+#    embedding = average_embedding
+#    return embedding
+#
+#
+#@app.route('/embeddings', methods=['POST'])
+#def embeddings():
+#    input_text_list = request.get_json().get('input')
+#    input_text      = ' '.join(map(str, input_text_list))
+#    token           = request.headers.get('Authorization').replace('Bearer ', '')
+#    embedding       = get_embedding(input_text, token)
+#    
+#    return {
+#        'data': [
+#            {
+#                'embedding': embedding,
+#                'index': 0,
+#                'object': 'embedding'
+#            }
+#        ],
+#        'model': 'text-embedding-ada-002',
+#        'object': 'list',
+#        'usage': {
+#            'prompt_tokens': None,
+#            'total_tokens': None
+#        }
+#    }
+#
 
 def main():
     asgi_app = WsgiToAsgi(app)