Instead of making your LLM fill in JSON schemas one tool call at a time, Dragen gives it a Python sandbox. The agent writes real code — loops, branches, error handling, multi-step reasoning — all in one shot. That's the CodeAct pattern.
Code runs in a Littrs sandbox — a Python-to-bytecode compiler and stack VM embedded directly in your process. No containers, no cloud sandboxing services, no exec(). Zero ambient capabilities: no filesystem, no network, no env vars. The only way sandboxed code reaches the outside world is through tools you explicitly provide.
Installation
pip install dragen
Quick Start
import dragen
agent = dragen.Agent("moonshotai/kimi-k2.5")
@agent.tool
def search(query: str) -> str:
"""Search the web for information."""
return f"Results for: {query}"
result = agent.run("Search for recent AI agent frameworks")
print(result)
Examples
Structured output with self-correction
Pass a schema and the agent retries until the output validates:
from pydantic import BaseModel
class Analysis(BaseModel):
summary: str
sentiment: str # positive, negative, neutral
confidence: float
agent = dragen.Agent("moonshotai/kimi-k2.5")
result = agent.run(
"Analyze the sentiment of: 'This product is amazing!'",
schema=Analysis.model_json_schema()
)
analysis = Analysis(**result)
Multi-agent pipeline with shared context
Agents pass typed data to each other through a shared Context:
from dragen import Agent, Context
ctx = Context()
# Planner researches and writes a plan
planner = Agent("moonshotai/kimi-k2.5").to_context(ctx, "plan")
planner.run("Create a research plan for: quantum computing trends")
# Writer reads the plan and produces content
writer = Agent("moonshotai/kimi-k2.5").from_context(ctx, "plan")
result = writer.run("Write a report based on the research plan")
Sandbox with tools, limits, and file access
sandbox = dragen.Sandbox(builtins=True)
sandbox.limit(max_instructions=50_000, max_recursion_depth=30)
sandbox.mount("data.csv", "./input/data.csv")
@sandbox.tool
def summarize(text: str) -> str:
"""Summarize text using an external API."""
return call_summary_api(text)
agent = dragen.Agent("moonshotai/kimi-k2.5", sandbox=sandbox)
result = agent.run("Read data.csv and summarize its contents")
Recursive Language Model (RLM)
Process inputs far beyond the context window — the long input lives in the sandbox as a variable and the agent writes code to slice and summarize it across iterations:
sandbox = dragen.Sandbox(builtins=True)
sandbox["document"] = very_long_text # e.g. 500K tokens
agent = dragen.Agent("moonshotai/kimi-k2.5", max_iterations=20, sandbox=sandbox)
result = agent.run("""
The variable `document` contains a very long research paper.
Extract all key findings, then synthesize them into a structured summary.
You can slice `document` with Python string indexing to read it in parts.
""")
Configuration
agent = dragen.Agent(
"moonshotai/kimi-k2.5",
max_iterations=10,
temperature=0.7,
max_tokens=4096,
system="You are a helpful assistant"
)
Event Callbacks
agent = dragen.Agent("moonshotai/kimi-k2.5")
@agent.on_code
def on_code(code):
print(f"Executing:\n{code}")
@agent.on_output
def on_output(output):
print(f"Output: {output}")
@agent.on_finish
def on_finish(result):
print(f"Done: {result}")
For the full feature reference, see DOCS.md. More examples in examples/.
License
Apache-2.0
Metadata
Release files for dragen 0.3.1
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Built distributions (wheels)
Total release size: 63.5 MB
Release files / dragen-0.3.1-cp313-cp313-win_amd64.whl
| Download URL | dragen-0.3.1-cp313-cp313-win_amd64.whl |
|---|---|
| Size | 4.2 MB |
| Tags | CPython 3.13 Windows x86-64 |
|
SHA-256 checksum How to use checksums |
43a1006801769cbf19d0472e094ba2251567e3023d40a5774a122bba19263232
|
|
BLAKE2b-256 checksum How to use checksums |
bfb0ee69f33af978af8a672d10ff0b12990ce961e344af9f56a55b603a86927f
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
Release files / dragen-0.3.1-cp313-cp313-manylinux_2_38_x86_64.whl
| Download URL | dragen-0.3.1-cp313-cp313-manylinux_2_38_x86_64.whl |
|---|---|
| Size | 7.1 MB |
| Tags | CPython 3.13 Linux glibc 2.38+ x86-64 |
|
SHA-256 checksum How to use checksums |
221fff01ce615f832f02758e43b90d9ff6d9039ac7875717a184b25c79c2a24a
|
|
BLAKE2b-256 checksum How to use checksums |
8a1125ef74a951bdc719c362c34978be3a72a1b1f44bf00225dd144a74dee4f2
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
Release files / dragen-0.3.1-cp313-cp313-macosx_11_0_arm64.whl
| Download URL | dragen-0.3.1-cp313-cp313-macosx_11_0_arm64.whl |
|---|---|
| Size | 4.5 MB |
| Tags | CPython 3.13 macOS 11.0+ ARM64 |
|
SHA-256 checksum How to use checksums |
cc2bd2e7c6cc8aa7ab128536da11c5ca711412755c158d7a9f348a9b3c7e86b1
|
|
BLAKE2b-256 checksum How to use checksums |
81b91087d679e3da19db9a681f479f61629a593747896ddafca24060ab466b67
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
Release files / dragen-0.3.1-cp312-cp312-win_amd64.whl
| Download URL | dragen-0.3.1-cp312-cp312-win_amd64.whl |
|---|---|
| Size | 4.2 MB |
| Tags | CPython 3.12 Windows x86-64 |
|
SHA-256 checksum How to use checksums |
893eae35a13169c67c915d69a0a0c145c8f6f4f34d86417b8d82e2e56e05cf31
|
|
BLAKE2b-256 checksum How to use checksums |
40b7b6cff46dcb159272919fd84ec300bcff2b653ee6f23a3be17a8244cdcde6
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
Release files / dragen-0.3.1-cp312-cp312-manylinux_2_38_x86_64.whl
| Download URL | dragen-0.3.1-cp312-cp312-manylinux_2_38_x86_64.whl |
|---|---|
| Size | 7.1 MB |
| Tags | CPython 3.12 Linux glibc 2.38+ x86-64 |
|
SHA-256 checksum How to use checksums |
6dea9c6c27d670c615c36cf4b3b7284d3685fd468034a9fe857e7ce9ba3ebd36
|
|
BLAKE2b-256 checksum How to use checksums |
b4701efca31285b12c48077fa4ec06e9ead3839b69d95838da319ae96f3a4720
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
Release files / dragen-0.3.1-cp312-cp312-macosx_11_0_arm64.whl
| Download URL | dragen-0.3.1-cp312-cp312-macosx_11_0_arm64.whl |
|---|---|
| Size | 4.5 MB |
| Tags | CPython 3.12 macOS 11.0+ ARM64 |
|
SHA-256 checksum How to use checksums |
bfe11ef3716d7621ffb8e358f864677b9e1b764a1a5b6f290eaee685a09c2f20
|
|
BLAKE2b-256 checksum How to use checksums |
5bdd1bf311ae142582ae09642450650e114ac95f578bbb5509851eca80a62bae
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
Release files / dragen-0.3.1-cp311-cp311-win_amd64.whl
| Download URL | dragen-0.3.1-cp311-cp311-win_amd64.whl |
|---|---|
| Size | 4.2 MB |
| Tags | CPython 3.11 Windows x86-64 |
|
SHA-256 checksum How to use checksums |
cc8d8c17c3e4f8319fa39fbbcf442652900db443aa32ffd8e5570dda23472acd
|
|
BLAKE2b-256 checksum How to use checksums |
39a931ca53fdc53cc24f9a1750494c5c65724edaa24016c659b92d3a2eb0a331
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
Release files / dragen-0.3.1-cp311-cp311-manylinux_2_38_x86_64.whl
| Download URL | dragen-0.3.1-cp311-cp311-manylinux_2_38_x86_64.whl |
|---|---|
| Size | 7.2 MB |
| Tags | CPython 3.11 Linux glibc 2.38+ x86-64 |
|
SHA-256 checksum How to use checksums |
6ffddcd3fba23f4e99c426939545d094d5fbb51fd906bd6e3aa962251eb9df3a
|
|
BLAKE2b-256 checksum How to use checksums |
9016c21edf5b303d21c3611f46c85fb50eb99850326b24223f542af0db26d3a6
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
Release files / dragen-0.3.1-cp311-cp311-macosx_11_0_arm64.whl
| Download URL | dragen-0.3.1-cp311-cp311-macosx_11_0_arm64.whl |
|---|---|
| Size | 4.5 MB |
| Tags | CPython 3.11 macOS 11.0+ ARM64 |
|
SHA-256 checksum How to use checksums |
3e0bc47a44e56354cd0c0b56999aab8e099ee093b633a63c3064facb4a45311f
|
|
BLAKE2b-256 checksum How to use checksums |
c0ee733019b7cdd07fef355c13629ecb2acc233c4d5670e039624abd6ab98991
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
Release files / dragen-0.3.1-cp310-cp310-win_amd64.whl
| Download URL | dragen-0.3.1-cp310-cp310-win_amd64.whl |
|---|---|
| Size | 4.2 MB |
| Tags | CPython 3.10 Windows x86-64 |
|
SHA-256 checksum How to use checksums |
db263d421680c4512ac145c18a3fc918955b23cca1021787fd93c4105f011850
|
|
BLAKE2b-256 checksum How to use checksums |
9b7c0051209df268c13614aaac4640fb7ba3277bbdb19eb5e1a3075c587fcea6
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
Release files / dragen-0.3.1-cp310-cp310-manylinux_2_38_x86_64.whl
| Download URL | dragen-0.3.1-cp310-cp310-manylinux_2_38_x86_64.whl |
|---|---|
| Size | 7.2 MB |
| Tags | CPython 3.10 Linux glibc 2.38+ x86-64 |
|
SHA-256 checksum How to use checksums |
a9987914268ed4194097a5013e8f180c74d11ad402115796e63607f549209d8a
|
|
BLAKE2b-256 checksum How to use checksums |
158e13b86325b33a6772b0ae74688efcc3cc29bf85664b8d54e2d01eb0fc71d4
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
Release files / dragen-0.3.1-cp310-cp310-macosx_11_0_arm64.whl
| Download URL | dragen-0.3.1-cp310-cp310-macosx_11_0_arm64.whl |
|---|---|
| Size | 4.5 MB |
| Tags | CPython 3.10 macOS 11.0+ ARM64 |
|
SHA-256 checksum How to use checksums |
1c1dcfc142e94e95f37a2cd7da8add0bbe082d883af781fa646bef61da46295c
|
|
BLAKE2b-256 checksum How to use checksums |
8ae1820f1bab17ba8fef04f147bad3b0ecc7bece1481237196f9c7079c599246
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
Security for dragen 0.3.1
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.