ReliabilityLane

Reliability patterns for AI agents. Match a failure symptom to a known pattern, plan a retry with backoff, verify work against a checklist, and resolve incidents with a playbook — all from a curated, deterministic dataset. No LLM guessing.

Why it exists

Agents fail in predictable ways: they retry the same broken call forever, trust a green check without verifying the end state, grade their own work, lose context, and let a language model do arithmetic. ReliabilityLane turns those failures into deterministic rules so agents fail safer, retry smarter, and prove their work.

Open dataset and CLI today. Hosted power behind them.

Install

pip install reliabilitylane

Quickstart

from reliabilitylane import ReliabilityLaneClient

client = ReliabilityLaneClient()  # reads TALOCODE_API_KEY, TALOCODE_BASE_URL

# Plan a retry for a 500
plan = client.retry_plan(status=500)
print(plan["strategy"]["name"], plan["nextDelayMs"])

# Match a failure symptom to a pattern
match = client.match(symptom="it keeps retrying the same call forever")
print(match["matches"][0]["pattern"]["name"])

# Verify a deployment
result = client.verify(
    area="deployment",
    evidence={"Deploy command exited with a success status": True},
)
print(result["verdict"], result["pending"])

No key needed for the local deterministic engine:

from reliabilitylane import match_failure, plan_retry, verify, incident_for

plan_retry(status=429)                      # rate-limit strategy, jittered delay
match_failure(symptom="retries forever")    # retry-hammering pattern
verify(area="code", evidence={"Lint passes": True})
incident_for(symptom="deploy failed")

Auth / env

• TALOCODE_API_KEY — hosted API key from StackLane Cloud
• TALOCODE_BASE_URL — defaults to https://api.talocode.site

API surface

Hosted endpoints (/v1/reliabilitylane/*)

Credits: patterns/retry/verify 1 cr, incident 2 cr, reads from dataset 0–1 cr.

Client methods

health, pricing, capabilities, patterns, retries, checklists, playbooks, antipatterns, match(symptom=..., error=..., category=...), retry_plan(status=..., code=..., message=...), verify(area=..., checklist=..., evidence={...}), incident(symptom=..., error=...).

Local engine

classify_error, match_failure, plan_retry, verify, incident_for, get_reliabilitylane_pricing, get_reliabilitylane_capabilities.

CLI

reliabilitylane match "it keeps retrying the same call forever"
reliabilitylane retry-plan --status 429
reliabilitylane verify --area deployment --evidence "Service responds"
reliabilitylane incident "deploy failed"
reliabilitylane antipatterns
reliabilitylane patterns
reliabilitylane retries
reliabilitylane checklists
reliabilitylane playbooks
reliabilitylane list
reliabilitylane pricing
reliabilitylane capabilities

Related packages

Talocode ecosystem

More: github.com/talocode · talocode.site · docs.talocode.site

License

MIT © Talocode