Skip to content

artifacts.md

Latest commit

 

History

History
147 lines (108 loc) · 3.27 KB

artifacts.md

File metadata and controls

147 lines (108 loc) · 3.27 KB

Artifacts

Artifacts provide persistent data storage across sessions. Use them to cache results, maintain state, or share data between workflows.

Basic Usage

# Save data
artifacts.save("analysis_results", {
    "repos_analyzed": 42,
    "findings": [...]
})

# Load data
data = artifacts.load("analysis_results")

# List all artifacts
all_artifacts = artifacts.list()

# Delete an artifact
artifacts.delete("old_data")

Storage Formats

Artifacts automatically handle serialization based on data type:

# JSON-serializable data (dicts, lists, primitives)
artifacts.save("config", {"api_key": "...", "timeout": 30})

# Binary data
artifacts.save("image", image_bytes)

# Text data
artifacts.save("report", "Analysis results: ...")

Use Cases

Caching API Responses

async def run(owner: str, repo: str) -> dict:
    cache_key = f"repo_{owner}_{repo}"

    # Check cache first
    cached = artifacts.load(cache_key)
    if cached:
        return cached

    # Fetch fresh data
    data = tools.curl.get(url=f"https://api.github.com/repos/{owner}/{repo}")

    # Cache for next time
    artifacts.save(cache_key, data)
    return data

Maintaining State

async def run(url: str) -> dict:
    # Load previous crawl state
    state = artifacts.load("crawl_state") or {"visited": [], "queue": []}

    if url in state["visited"]:
        return {"status": "already_crawled"}

    # Process URL
    content = tools.fetch(url=url)
    state["visited"].append(url)

    # Save updated state
    artifacts.save("crawl_state", state)
    return {"status": "success", "content": content}

Sharing Data Between Workflows

# Skill 1: Collect data
async def run(sources: list) -> dict:
    results = [fetch_source(s) for s in sources]
    artifacts.save("collected_data", results)
    return {"count": len(results)}

# Skill 2: Analyze data
async def run() -> dict:
    data = artifacts.load("collected_data")
    analysis = analyze(data)
    artifacts.save("analysis_report", analysis)
    return analysis

Storage Backends

Artifacts are stored according to your storage backend:

FileStorage:

  • Stored in {base_path}/artifacts/ directory
  • Each artifact is a separate file
  • JSON data stored as .json, binary as raw files

RedisStorage:

  • Stored as Redis keys with configured prefix
  • Automatic expiration support (if configured)
  • Shared across all agent instances

Best Practices

Use descriptive names:

# Good
artifacts.save("github_repos_2024_analysis", data)

# Bad
artifacts.save("data1", data)

Clean up old artifacts:

# Remove artifacts you no longer need
artifacts.delete("temp_processing_results")

Consider data size:

  • Artifacts are loaded into memory when accessed
  • Large datasets may impact performance
  • Consider pagination or chunking for large data

Metadata

Artifacts automatically track metadata:

artifacts_list = artifacts.list()
for artifact in artifacts_list:
    print(f"{artifact['name']}: {artifact['created_at']}")

Limitations

  • No versioning (saving with same name overwrites)
  • No automatic expiration (except with Redis TTL configuration)
  • No search/filtering beyond list all
  • No access control (all agents with same storage can access all artifacts)