Tencent Cloud CLS (Log Service) — Search & Analysis

Search and run CQL / SQL analytics over Tencent Cloud CLS log topics.

Setup: See tencentcloud authentication. The SDK reads TENCENTCLOUD_SECRET_ID / TENCENTCLOUD_SECRET_KEY / TENCENTCLOUD_REGION from the environment.

Companion skill: Use tencentcloud-cls-alarm for alarm policy / notice group / shield management. This skill is only about searching log content.

CLI (preferred)

The skill ships scripts/cls.py — a self-contained CLI for the most common operations.

# $SKILL_DIR can point at another skill loaded this turn — anchor on our own
# script (re-run this at the top of every fresh-shell Bash block).
CLS="$SKILL_DIR/scripts/cls.py"; [ -f "$CLS" ] || CLS=$(find /tmp -maxdepth 8 -path '*/skills/*/scripts/cls.py' 2>/dev/null | head -1)
[ -f "$CLS" ] || { echo "tencentcloud-cls script not found (SKILL_DIR=$SKILL_DIR)" >&2; exit 1; }

python3 $CLS topics                                              # list topics
python3 $CLS search --topic <topic-id> --query 'level:ERROR' --time 1h
python3 $CLS search --topic <topic-id> --trace-id <uuid>
python3 $CLS search --topic <topic-id> --time 1d \
    --query '* | SELECT api_name, count(*) AS cnt GROUP BY api_name ORDER BY cnt DESC LIMIT 20' \
    --format json

--time accepts 30m / 1h / 6h / 1d / 7d. --query is CQL by default; pass --lucene to switch dialect. Append | SELECT ... GROUP BY ... to a query for SQL analytics.

For anything beyond what the CLI exposes (custom field projections, raw paginated walks, Context-based tailing) drop down to the SDK calls below.

When to Use

• Find recent errors / 5xx responses for a service
• Look up a single request by trace ID across multiple topics
• Run CQL filters (status_code:>=500 AND service:"openai")
• Run SQL analytics (SELECT api_name, COUNT(*) GROUP BY api_name)
• Chain trace logs to reconstruct a request lifecycle
• Extract specific fields for billing / audit reports

Dependencies

pip install tencentcloud-sdk-python

Quick start

import os
import json
from tencentcloud.common import credential
from tencentcloud.cls.v20201016 import cls_client, models

cred = credential.EnvironmentVariableCredential().get_credential()
client = cls_client.ClsClient(cred, os.environ["TENCENTCLOUD_REGION"])

Workflows

Discover topics

req = models.DescribeTopicsRequest()
resp = client.DescribeTopics(req)
for t in resp.Topics:
    print(t.TopicId, t.TopicName, t.LogsetId)

Tip: ask the user for the topic ID up front. Topic IDs look like 751a7350-dc5d-41c3-a6ca-c178bae05807 and aren't guessable from a service name.

Search logs (CQL)

import time

req = models.SearchLogRequest()
req.TopicId = "<topic-id>"
req.From = int((time.time() - 3600) * 1000)   # 1 hour ago, ms epoch
req.To = int(time.time() * 1000)
req.Query = 'status_code:>=500 AND service:"openai"'
req.Limit = 100
req.Sort = "desc"
req.SyntaxRule = 1                             # 1 = CQL, 0 = Lucene

resp = client.SearchLog(req)
for line in resp.Results:
    fields = {f.Key: f.Value for f in line.LogJson and []}  # see below
    print(line.Time, line.PkgLogId, fields.get("status_code"), fields.get("api_name"))

CLS hands you Results with LogJson as a JSON string per record. Decode with json.loads(line.LogJson) to get a flat dict of all the indexed fields the topic stores.

Look up a trace ID across multiple topics

TRACE_ID = "34341776-0835-422a-956d-ac8d5b404db1"
TOPICS = {
    "trace": "<trace-topic-id>",
    "api-usages": "<api-usages-topic-id>",
}

for label, topic_id in TOPICS.items():
    req = models.SearchLogRequest()
    req.TopicId = topic_id
    req.From = int((time.time() - 86400) * 1000)
    req.To = int(time.time() * 1000)
    req.Query = f'trace_id:"{TRACE_ID}"'
    req.Limit = 100
    req.SyntaxRule = 1
    resp = client.SearchLog(req)
    print(f"--- {label}: {len(resp.Results)} records ---")
    for line in resp.Results:
        print(line.Time, line.LogJson)

SQL analytics

CLS supports a SQL-on-logs subset for aggregation. Append | <SQL> to a CQL filter:

req.Query = '* | SELECT api_name, count(*) AS cnt GROUP BY api_name ORDER BY cnt DESC LIMIT 20'
req.SyntaxRule = 1
resp = client.SearchLog(req)
# When the query has a `|`, results land in resp.Analysis (rows of column→value)
for row in resp.AnalysisResults or []:
    print(row)

Tail recent logs

# Poll every 5s for new lines after the last cursor
last_cursor = None
while True:
    req = models.SearchLogRequest()
    req.TopicId = topic_id
    req.From = int((time.time() - 30) * 1000)
    req.To = int(time.time() * 1000)
    req.Query = 'level:ERROR'
    req.Limit = 100
    req.Sort = "asc"
    req.SyntaxRule = 1
    if last_cursor:
        req.Context = last_cursor
    resp = client.SearchLog(req)
    for line in resp.Results:
        print(line.Time, line.LogJson)
    last_cursor = resp.Context
    time.sleep(5)

CQL cheatsheet

Pagination

CLS returns up to 1000 records per call. For larger result sets, iterate with Context:

all_records = []
context = None
while True:
    req = models.SearchLogRequest()
    req.TopicId = topic_id
    req.From = ...
    req.To = ...
    req.Query = '...'
    req.Limit = 1000
    req.SyntaxRule = 1
    if context:
        req.Context = context
    resp = client.SearchLog(req)
    all_records.extend(resp.Results)
    context = resp.Context
    if not context or len(resp.Results) < 1000:
        break

Error patterns

Console links

• CLS console: https://console.cloud.tencent.com/cls/topic
• CQL syntax: https://www.tencentcloud.com/document/product/614/47044
• SQL analytics syntax: https://www.tencentcloud.com/document/product/614/58978