🧬 AgentDNA — Sentry for AI Agents
One-line observability for any Python agent. No framework required. No API key. No network calls.
pip install agentdna
Quick Start
from agentdna import observe, get_stats
@observe
def my_agent(prompt):
# your existing agent code
return llm.call(prompt)
my_agent("hello world")
# View stats (persists across restarts)
print(get_stats())
That's it. One decorator. Full observability.
What You Get
📊 Stats: my_agent
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Health: ✅ Healthy
Total calls: 42
Success rate: 97.6%
Failed calls: 1
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Avg latency: 1250.5 ms
P50 latency: 980.0 ms
P95 latency: 2100.0 ms
P99 latency: 3500.0 ms
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Errors:
Timeout: 1
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Features
- 📊 Call tracking — total calls, success/failure rates
- ⏱️ Latency — avg, p50, p95, p99 percentiles
- ❌ Error tracking — error types and frequency
- 💾 SQLite persistence — data survives restarts
- 🔒 100% local — no network calls, no API key
- 🐍 Sync & async — works with both
- 🖥️ CLI included —
agentdna stats
CLI
agentdna stats # overview of all observed functions
agentdna stats my_agent # detailed view
agentdna stats --export json # export as JSON
agentdna stats --export csv # export as CSV
agentdna stats --reset # clear all data
Usage
Basic
from agentdna import observe
@observe
def my_agent(prompt):
return llm.call(prompt)
With Options
@observe(name="transcriber", tags={"version": "2.0", "model": "whisper"})
def transcribe(audio_path):
return whisper.transcribe(audio_path)
Async Support
@observe
async def my_async_agent(prompt):
result = await llm.acall(prompt)
return result
Get Stats Programmatically
from agentdna import get_stats
stats = get_stats("my_agent")
print(f"Success rate: {stats['success_rate']:.1%}")
print(f"P95 latency: {stats['p95_latency_ms']:.0f}ms")
# All functions
all_stats = get_stats()
for name, s in all_stats.items():
print(f"{name}: {s['total_calls']} calls")
Export Stats
from agentdna import export_stats
json_str = export_stats(format="json")
csv_str = export_stats(format="csv")
How It Works
- Decorator wraps your function
- Each call logs: timestamp, success/failure, latency, input/output sizes, errors
- Data batches to
~/.agentdna/observe.db(SQLite, WAL mode) - Read stats anytime via
get_stats()oragentdna stats
No dependencies beyond Python stdlib. click is only needed for the CLI.
Data Location
Default: ~/.agentdna/observe.db
Custom path:
import os
os.environ["AGENTDNA_DB_PATH"] = "/path/to/my/observe.db"
Why AgentDNA?
| Without AgentDNA | With AgentDNA |
|---|---|
print("done!") |
Track every call with latency + errors |
| Hope nothing breaks | Know exactly when and what breaks |
| Debug blind | P95 latency, error breakdown, trends |
| Lose data on restart | SQLite persistence |
Installation
pip install agentdna
Optional features:
pip install agentdna[discovery] # agent discovery (httpx, PyYAML)
pip install agentdna[server] # run registry server (FastAPI)
License
Apache 2.0
Metadata
Release files for agentdna-sdk 0.2.1
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 |
|---|---|---|---|---|
| agentdna_sdk-0.2.1.tar.gz | 26.9 kB | Without provenance | Details for agentdna_sdk-0.2.1.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| agentdna_sdk-0.2.1-py3-none-any.whl | Python 3 | none | any | Details for agentdna_sdk-0.2.1-py3-none-any.whl |
Total release size: 61.4 kB
Release files / agentdna_sdk-0.2.1.tar.gz
| Download URL | agentdna_sdk-0.2.1.tar.gz |
|---|---|
| Size | 26.9 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
2b7489fccd45b352f038fe3a5896998ba4e16cba09dd118ff9aa510b07a5eb2e
|
|
BLAKE2b-256 checksum How to use checksums |
2d5b1b952eebadf79a30322352308763f13dbef2434f51ea0a4b78da789fe16d
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.12.3
|
Release files / agentdna_sdk-0.2.1-py3-none-any.whl
| Download URL | agentdna_sdk-0.2.1-py3-none-any.whl |
|---|---|
| Size | 34.4 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
f9208ca111cf5c6baf93fbafdeb0687774b4a032acf668d7ab8793d202c16fcf
|
|
BLAKE2b-256 checksum How to use checksums |
a59c5d53a0277d47a6926ee4c5071924b2843a1bcce8c71535991805fcdb73bd
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.12.3
|
Security for agentdna-sdk 0.2.1
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.