Migrate to langfuse

alttexter.py

@@ -2,38 +2,40 @@
 import logging
 import mimetypes
 import os
-import time
 from typing import List, Optional, Tuple
 
 import cairosvg
-from langchain import callbacks
-from langchain.callbacks.tracers.langchain import wait_for_all_tracers
 from langchain.output_parsers import PydanticOutputParser
 from langchain.prompts import ChatPromptTemplate
 from langchain_core.messages import HumanMessage, SystemMessage
 from langchain_openai import AzureChatOpenAI, ChatOpenAI
-from langsmith import Client
+from langfuse.callback import CallbackHandler
 
 from schema import AlttexterResponse, ImageAltText
 
+langfuse_handler = CallbackHandler(
+    public_key=os.getenv('LANGFUSE_PUBLIC_KEY'),
+    secret_key=os.getenv('LANGFUSE_SECRET_KEY'),
+    host=os.getenv('LANGFUSE_HOST')
+)
+
 
 def determine_llm() -> ChatOpenAI:
     """Determine which LLM to use based on environment variable."""
     model_env = os.getenv("ALTTEXTER_MODEL")
     if model_env == "openai":
         return ChatOpenAI(verbose=True,
                           temperature=0,
-                          model="gpt-4-vision-preview",
-                          max_tokens=4096)
+                          model="gpt-4o")
     elif model_env == "openai_azure":
         return AzureChatOpenAI(verbose=True,
                                temperature=0, openai_api_version="2024-02-15-preview",
                                azure_deployment=os.getenv("AZURE_OPENAI_DEPLOYMENT"),
-                               model="vision-preview",
-                               max_tokens=4096)
+                               model="gpt-4o")
     else:
         raise ValueError(f"Unsupported model specified: {model_env}")
 
+
 def svg_to_png_base64(svg_data):
     """
     Converts SVG data to PNG and returns the base64 encoded PNG image.
@@ -48,6 +50,7 @@ def svg_to_png_base64(svg_data):
 
     return base64.b64encode(png_data).decode('utf-8')
 
+
 def alttexter(input_text: str, images: dict, image_urls: List[str]) -> Tuple[List[ImageAltText], Optional[str]]:
     """
     Processes input text and images to generate alt text and title attributes.
@@ -122,29 +125,11 @@ def alttexter(input_text: str, images: dict, image_urls: List[str]) -> Tuple[Lis
     alttexts = None
     run_url = None
 
-    tracing_enabled = os.getenv("LANGCHAIN_TRACING_V2", "").lower() == "true"
-    if tracing_enabled:
-        client = Client()
-        try:
-            with callbacks.collect_runs() as cb:
-                alttexts = llm.invoke(messages.format_messages())
-
-                # Ensure that all tracers complete their execution
-                wait_for_all_tracers()
-
-                if alttexts:
-                    # Get public URL for run
-                    run_id = cb.traced_runs[0].id
-                    time.sleep(2)
-                    client.share_run(run_id)
-                    run_url = client.read_run_shared_link(run_id)
-        except Exception as e:
-            logging.error(f"Error during LLM invocation with tracing: {str(e)}")
+    if os.getenv("LANGFUSE_TRACING", "False"):
+        alttexts = llm.invoke(messages.format_messages(), config={"callbacks": [langfuse_handler]})
+        run_url = str(langfuse_handler.get_trace_url())
     else:
-        try:
-            alttexts = llm.invoke(messages.format_messages())
-        except Exception as e:
-            logging.error(f"Error during LLM invocation without tracing: {str(e)}")
+        alttexts = llm.invoke(messages.format_messages())
 
     if alttexts:
         try:

client-example.py

@@ -1,13 +1,14 @@
+import argparse
+import base64
+import getpass
 import json
+import logging
 import os
 import re
-import base64
-import logging
-import requests
-import getpass
-import argparse
 from datetime import datetime
 
+import requests
+
 logging.basicConfig(level=logging.INFO,
                     format='%(levelname)s [%(asctime)s] %(message)s',
                     datefmt='%d-%m-%Y %H:%M:%S')
@@ -67,7 +68,7 @@ def log_payload_summary(encoded_images, image_urls):
         logging.info(f"Image URLs: {image_urls}")
 
 
-def send_file_to_api(md_content, encoded_images, image_urls, url, token, full_payload):
+def send_file_to_api(md_content, encoded_images, image_urls, url, token, full_payload, verify_ssl=True):
     if full_payload:
         log_full_payload(md_content, encoded_images, image_urls)
     else:
@@ -86,7 +87,20 @@ def send_file_to_api(md_content, encoded_images, image_urls, url, token, full_pa
     }
 
     logging.info("Sending payload to alttexter...")
-    response = requests.post(url, headers=headers, data=actual_payload, timeout=120)
+    try:
+        response = requests.post(url, headers=headers, data=actual_payload, timeout=120, verify=verify_ssl)
+        response.raise_for_status()  # Raises an HTTPError for bad responses
+    except requests.exceptions.SSLError as ssl_err:
+        if verify_ssl:
+            logging.error(f"SSL Error occurred: {ssl_err}")
+            raise
+        else:
+            logging.warning("SSL verification is disabled. Proceeding with insecure request.")
+            response = requests.post(url, headers=headers, data=actual_payload, timeout=120, verify=False)
+            response.raise_for_status()
+    except requests.exceptions.RequestException as req_err:
+        logging.error(f"An error occurred while sending the request: {req_err}")
+        raise
 
     timestamp = datetime.now().strftime('%d-%m-%Y %H:%M:%S')
     logging.info(f"Response received at {timestamp}")
@@ -98,6 +112,7 @@ def send_file_to_api(md_content, encoded_images, image_urls, url, token, full_pa
     parser = argparse.ArgumentParser(description="Send markdown file to alttexter")
     parser.add_argument("md_file_path", help="Path to file containing markdown formatted text.")
     parser.add_argument("--full", action="store_true", help="Log the full payload instead of the summary")
+    parser.add_argument("--no-verify-ssl", action="store_true", help="Disable SSL certificate verification")
 
     args = parser.parse_args()
 
@@ -117,5 +132,10 @@ def send_file_to_api(md_content, encoded_images, image_urls, url, token, full_pa
     local_images, image_urls = extract_images_from_markdown(md_content)
     encode_local_images(local_images, base_dir)
 
-    response = send_file_to_api(md_content, local_images, image_urls, url, token, args.full)
-    print(response)
+    verify_ssl = not args.no_verify_ssl
+    try:
+        response = send_file_to_api(md_content, local_images, image_urls, url, token, args.full, verify_ssl)
+        print(response)
+    except requests.exceptions.RequestException as e:
+        logging.error(f"Error occurred: {e}")
+        print(f"Failed to get a response from the server: {e}")

docker-compose.yml

@@ -11,10 +11,10 @@ services:
     volumes:
       - ${ALTTEXTER_CERTS_DIR}:/certs
     environment:
-      - LANGCHAIN_TRACING_V2
-      - LANGCHAIN_ENDPOINT
-      - LANGCHAIN_PROJECT
-      - LANGCHAIN_API_KEY
+      - LANGFUSE_TRACING
+      - LANGFUSE_PUBLIC_KEY
+      - LANGFUSE_SECRET_KEY
+      - LANGFUSE_HOST
       - ALTTEXTER_MODEL
       - OPENAI_API_KEY
       - AZURE_OPENAI_ENDPOINT

requirements.txt

@@ -5,4 +5,5 @@ pydantic==1.10.12
 uvicorn==0.27.1
 tiktoken==0.5.2
 nbformat==5.9.2
-cairosvg==2.7.0
+cairosvg==2.7.0
+langfuse==2.36.1