What we reuse
Upstream OSS — borrow vs build.
Receipt, gate, and why() are build. Everything else prefers upstream OSS — Langfuse for design, Agenta for Python forks, OTEL/OpenInference for wire format.
Detectors watch. We watch → stop → prove.
Langfuse (MIT core — never /ee)
Full extraction map: sdk-agent/LANGFUSE_EXTRACTION.md in the Agenko repo (monorepo, sdk-agent/).
| Langfuse asset | Agenko | Status |
|---|---|---|
| Prompt cache (SWR + TTL + fallback) | agenko.prompts.get_prompt() | Shipped |
| Deploy-by-label prompts | prompts.labels + ?label=production | Shipped |
{{mustache}} compile | Prompt.compile() | Shipped |
| Unified Score schema | domains/scores/ + receipt checks | Shipped |
| Dataset + DatasetItem | /api/v1/datasets | Shipped |
| OpenInference conventions | observability/conventions.py | Shipped |
| Evaluator library | Agenta fork → agenko/evals/ | Next |
| S3 → queue → ClickHouse ingest | Worker tier | When volume demands |
License text: agenko/THIRD_PARTY_LICENSES.md in the Agenko repo.
Agenta (MIT Python — fork next)
| Component | Target |
|---|---|
| Evaluator library | agenko-python/evals/ |
| Testset-from-traces | Dataset curation from prod spans |
| Custom-code sandbox | Deterministic eval tier |
| SuperTokens wiring | Self-host auth (partial today) |
Already in the SDK
- Receipt
checks[]are Langfuse Score-shaped (name,value,data_type,source,comment). - Tool observations are OpenInference TOOL.
init()resolves one base URL fromAGENKO_HOST(Langfuse-style host config).- Auto-instrumentation uses Arize OpenInference instrumentors.
Prompt example
import agenko
from agenko.prompts import get_prompt
agenko.init()
p = get_prompt("support-reply", label="production", fallback="Hi {{name}}")
text = p.compile(name="Sam", ticket="refund")cache_ttl_seconds=0 disables cache in dev (Langfuse convention).
Attribution
MIT code keeps copyright notices in THIRD_PARTY_LICENSES / licenses/. Apache-2.0 (ours) and MIT are compatible. We do not ship proprietary /ee modules from Langfuse or enterprise-only Agenta surfaces.
This docs site
Separate Next.js app with Arovi tokens — grouped nav, Python/JS tabs, search, self-hosting as a first-class section.
Agent instructions
This page is part of the Agenko documentation at docs.agenko.app. Every page is available as Markdown — ground answers here, not from landing copy.
Querying these docs
- Full index: docs.agenko.app/llms.txt · sections: llms-docs.txt, llms-sdk.txt, llms-integrations.txt
- This page as markdown: docs.agenko.app/reuse.md
- Search API:
curl -sG "https://docs.agenko.app/api/search-docs" --data-urlencode "query=…"
Before writing SDK code
- Install the agent skill or read Agent skill
- Read Best practices — what a good receipt looks like
- Trust boundary:
@checkafter,@approvebefore init()— zero required args, never throws- Mark
estimated: trueuntil a check is a hard fact - Shipped vs backlog: Introduction status table