Extracting Trace Logs for Specific Traces Using Langfuse API#

Background#

During development, I needed to batch retrieve logs for specific traces, but Langfuse’s UI doesn’t support batch downloading specific traces as JSON files. Here’s a brief introduction on how to achieve this using the Langfuse API.

Obtaining sessionId#

A sessionId is required to retrieve all traces corresponding to that particular session batch. The method is simple: in the Langfuse UI interface, locate the sessionId column in the traces section and copy it out.

Obtaining Langfuse Keys and BASE_URL#

In the Langfuse UI’s Settings, find API Keys, create an API key. You’ll obtain two keys: PUBLIC_KEY and SECRET_KEY, both of which need to be recorded.

If your Langfuse is on the public network, the BASE_URL is https://api.langfuse.com/api/public/. If your Langfuse is local, the BASE_URL is http://YOUR_HOST/api/public/.

Relevant Code#

Finally, write the sessionId, PUBLIC_KEY, SECRET_KEY, and BASE_URL into langfuse_api.py, then run langfuse_api.py to obtain the logs for the specified traces.

1
import requests
2
import json
3

4
# Configuration
5
PUBLIC_KEY = ""
6
SECRET_KEY = ""
7
BASE_URL = "http://10.65.171.100:33000/api/public/"
8

9
# Target message_id to filter
10
TARGET_MESSAGE_ID = "17400353"
11
PAGE_SIZE=100
12

13
# Store results
14
results = 0
15
page = 1
16
has_more = True
17
target_date = "2025-08-06"
18
hour = 18
19

20

21
while has_more:
22
    print(f"📌 Fetching page {page} ({PAGE_SIZE} items per page)...")
23

24
    url = f"{BASE_URL}/traces"
25
    response = requests.get(
26
        url,
27
        auth=(PUBLIC_KEY, SECRET_KEY),
28
        params={
29
            "page":page,
30
            "limit":PAGE_SIZE,
31
            "sessionId":"test_cases_fivedoctors_20250806185514"
32
        },
33
    )
34
    res = response.json()
35

36

37
    if response.status_code != 200:
38
        print(f"❌ Request failed: {response.status_code} - {response.text}")
39
        break
40

41
    try:
42
        data = response.json()
43
    except json.JSONDecodeError:
44
        print("❌ Response content is not JSON format, possibly network issue or address error")
45
        print("Response content:", response.text)
46
        break
47

48
    if isinstance(data, list):
49
        traces = data
50
        has_more = len(traces) == PAGE_SIZE  # If exactly full page, there might be more
51
    else:
52
        traces = data.get("data", [])
53
        current_page_size = len(traces)
54
        has_more = current_page_size == PAGE_SIZE
55

56
    for trace in traces:
57
        metadata = trace.get("metadata") or {}
58
        id = metadata["message_id"]
59
        with open(f'./result/{id}.json',"w",encoding="utf-8") as f:
60
            json.dump(trace, f, indent=2, ensure_ascii=False)
61
            results+=1
62

63

64
    if len(traces) < PAGE_SIZE:
65
        has_more = False
66

67
    page += 1