SwarmSleuth (DevSwarm)
Local-first, evidence-backed incident triage. Apache 2.0 licensed, alpha release. Accept sanitized telemetry, fingerprint/deduplicate it, and generate a structured bug draft with exact evidence excerpts, cited observations, explicit unknowns and review next steps. No model key or cloud account required. No automatic GitHub writes or code execution.
Installation
The SwarmSleuth PyPI release is being prepared. You can install now from the public GitHub repository:
python -m pip install "git+https://github.com/rishiparpyaninortheastern/devswarm.git"
devswarm demo
This requires Git and Python 3.11+. Alternatively, install the prepared local wheel:
python -m pip install ./swarmsleuth-0.1.0-py3-none-any.whl
After the first PyPI release, the public command will be pip install swarmsleuth.
Use Python 3.11+. The base package requires Pydantic; optional integrations are separate:
# After publication, or use the matching wheel path with extras before publication:
pip install 'swarmsleuth[api]' # FastAPI + Uvicorn
pip install 'swarmsleuth[mcp]' # FastMCP tool servers
pip install 'swarmsleuth[azure]' # Azure Functions entry point
pip install 'swarmsleuth[all]' # All optional integrations
Offline quick start
devswarm --db ./incidents.db demo
devswarm sample > incident.json
devswarm --db ./incidents.db ingest incident.json
devswarm --db ./incidents.db worker --once
devswarm --db ./incidents.db show FINGERPRINT
devswarm --db ./incidents.db export FINGERPRINT
# After fixing a processing failure:
devswarm --db ./incidents.db retry FINGERPRINT
demo uses a bundled fixture, stores it locally and creates a reviewable draft. Repeated intake
increments its occurrence count. The fingerprint is returned by ingest and demo.
python -m devswarm supports the same commands. DEVSWARM_DB sets the database default;
--db overrides it. The fallback is .data/devswarm.db in the current working directory.
Operational events go to stderr; command results are JSON on stdout.
Python API
from devswarm import DevSwarm, Incident
swarm = DevSwarm("./incidents.db")
incident = Incident(
service="checkout",
environment="local",
exception_type="TimeoutError",
message="Inventory request timed out",
stack=["checkout.py:42 in reserve_inventory"],
occurred_at="2026-10-19T18:00:00Z",
sanitized=True,
)
receipt = swarm.ingest(incident)
swarm.process_next()
result = swarm.get(receipt.fingerprint)
print(result["draft"])
Optional integrations
With the corresponding extra installed:
devswarm --db ./incidents.db serve --port 8000
devswarm --db ./incidents.db worker --watch
devswarm --db ./incidents.db mcp evidence
devswarm mcp drafts
devswarm --db ./incidents.db init-functions ./functions-host
devswarm init-azure ./azure-foundation
The API binds to loopback. MCP uses stdio. init-functions creates host files for Core Tools;
install Core Tools and Azurite separately, activate the same Python environment, start Azurite,
then run func start in the generated host directory. init-azure exports a Terraform foundation;
it does not provision resources or provide cloud database/queue adapters. Initializers refuse
to overwrite existing files. Docker, Terraform, Core Tools and Azurite are external tools,
not binaries bundled into the Python package.
Security and limitations
Submit only upstream-sanitized telemetry. Defensive redaction is best effort, not comprehensive DLP. SQLite is plaintext and intended for local use. Root cause and exact reproduction remain unverified; drafts require human review. Fingerprint v2 normalizes durations, explicit request IDs, source line numbers and UUIDs while preserving HTTP/error codes and other numeric values. It is heuristic. Legacy alpha fingerprints remain readable but do not deduplicate with newly generated v2 keys; no automatic migration is provided. Forge is disabled and Courier exports JSON only. This alpha has no production authentication or hosted Azure deployment.
Processing failures
show and API reads include processing attempts, a safe error code and the next retry time
(Unix seconds). Failed work retries with short exponential backoff, stops automatically after three
failed attempts, and becomes failed. One bad incident does not block other eligible work.
Inspect/fix the problem, then use retry FINGERPRINT to reset pending/failed work; completed drafts
are never reset. Provider exception text is not stored or printed. Existing local databases receive
an additive retry-column migration that preserves records.
Credits
See credits and provenance for dependencies, related projects, and clarification about the separate product sharing this name.
Metadata
Release files for swarmsleuth 0.1.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| swarmsleuth-0.1.0.tar.gz | 58.4 kB | With provenance | Details for swarmsleuth-0.1.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| swarmsleuth-0.1.0-py3-none-any.whl | Python 3 | none | any | Details for swarmsleuth-0.1.0-py3-none-any.whl |
Total release size: 86.3 kB
Release files / swarmsleuth-0.1.0.tar.gz
| Download URL | swarmsleuth-0.1.0.tar.gz |
|---|---|
| Size | 58.4 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
063a586c2623d0c91957543423086fa0734822ef29a8c93faa7cafd632327314
|
|
BLAKE2b-256 checksum How to use checksums |
d881cb858629c19f9da699a97b815ab58814697bdc0faabb72b5004aa059b2fe
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
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.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 1, 2026.
Transparency logRelease files / swarmsleuth-0.1.0-py3-none-any.whl
| Download URL | swarmsleuth-0.1.0-py3-none-any.whl |
|---|---|
| Size | 27.9 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
40fb4643501e53669936ca2e5ea6b13dda506d4ccdf707079a0a28ce52700d56
|
|
BLAKE2b-256 checksum How to use checksums |
83d8dc0a1517d6b12cd2bd83a0fa19ee7fa299fab7a5771047f6e7f4914bcc9b
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
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.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 1, 2026.
Transparency logSecurity for swarmsleuth 0.1.0
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.
Full provenance
Both files in this release have been uploaded with provenance metadata.
Attestation for swarmsleuth-0.1.0.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 1, 2026.
Transparency logA 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.