llm-taskbench/docker-compose.cli.yml at main · knightsri/llm-taskbench · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
services:
  taskbench-cli:
    build:
      context: .
      dockerfile: Dockerfile
    env_file:
      - .env
    environment:
      - OPENROUTER_API_KEY=${OPENROUTER_API_KEY}
      - GENERAL_TASK_LLM=${GENERAL_TASK_LLM:-anthropic/claude-sonnet-4.5}
      - TASKBENCH_MAX_CONCURRENCY=${TASKBENCH_MAX_CONCURRENCY:-5}
      - TASKBENCH_USE_GENERATION_LOOKUP=${TASKBENCH_USE_GENERATION_LOOKUP:-true}
      - TASKBENCH_MAX_TOKENS=${TASKBENCH_MAX_TOKENS:-4000}
      - TASKBENCH_TEMPERATURE=${TASKBENCH_TEMPERATURE:-0.7}
      - TASKBENCH_CHUNK_CHARS=${TASKBENCH_CHUNK_CHARS:-20000}
      - TASKBENCH_CHUNK_OVERLAP=${TASKBENCH_CHUNK_OVERLAP:-500}
      - TASKBENCH_TIMEOUT=${TASKBENCH_TIMEOUT:-120.0}
      - TASKBENCH_RATE_LIMIT=${TASKBENCH_RATE_LIMIT:-60}
      - TASKBENCH_RETRY_MAX_DELAY=${TASKBENCH_RETRY_MAX_DELAY:-60.0}
      - TASKBENCH_JUDGE_MAX_TOKENS=${TASKBENCH_JUDGE_MAX_TOKENS:-2000}
      - TASKBENCH_INPUT_PREVIEW_LEN=${TASKBENCH_INPUT_PREVIEW_LEN:-5000}
    volumes:
      - ./tasks:/app/tasks
      - ./config:/app/config
      - ./results:/app/results
      - ./tests/fixtures:/app/tests/fixtures
      - ./sample-usecases:/app/sample-usecases
      - ./usecases:/app/usecases
      - ./.cache:/app/.cache
    working_dir: /app
    entrypoint: ["taskbench"]
    command: ["--help"]

# Usage Examples:
#
#   Build the CLI image:
#     docker compose -f docker-compose.cli.yml build
#
#   List available use cases:
#     docker compose -f docker-compose.cli.yml run --rm taskbench-cli list-usecases
#
#   Run a folder-based use case (recommended):
#     docker compose -f docker-compose.cli.yml run --rm taskbench-cli run \
#       sample-usecases/00-lecture-concept-extraction \
#       --models anthropic/claude-sonnet-4,openai/gpt-4o \
#       --skip-judge
#
#   Run with full judge evaluation:
#     docker compose -f docker-compose.cli.yml run --rm taskbench-cli run \
#       sample-usecases/00-lecture-concept-extraction \
#       --models anthropic/claude-sonnet-4
#
#   Legacy YAML-based evaluation:
#     docker compose -f docker-compose.cli.yml run --rm taskbench-cli evaluate \
#       tasks/lecture_analysis.yaml \
#       --models anthropic/claude-sonnet-4,openai/gpt-4o \
#       --input-file tests/fixtures/sample_transcript.txt
#
# Environment Variables:
#   Required: OPENROUTER_API_KEY (set in .env file)
#   See CLAUDE.md for all configurable environment variables