traceAI-typesafe
OpenTelemetry instrumentation for the TypeSafe AI Python SDK
(typesafe-sdk), the client for Jev's System One decision API (POST /v1/systemone).
Every TypeSafeClient.system_one(...) and AsyncTypeSafeClient.system_one(...) call becomes
one EVALUATOR span in Future AGI, however many questions it asks. Each question becomes
typesafe.question.<name>.* attributes:
| Answer type | Attributes |
|---|---|
noul |
type, noul (probability of yes, a number, never Pass/Fail) |
choice |
type, choice (the label), confidence |
score |
type, score (expected score, a number), confidence |
box |
type, box.top, box.bottom, box.left, box.right, confidence when not null |
The span also carries gen_ai.provider.name=typesafe, gen_ai.request.model (as you asked
for it), gen_ai.response.model plus typesafe.response.model when the API resolved an
alias, gen_ai.usage.input_tokens / output_tokens, typesafe.request_id, server.address
and typesafe.question_count. Probability maps, score legends and image bytes are never
recorded. The API returns no explanation, so the span has none.
This package traces calls your code makes to TypeSafe. It does not add Jev as a judge model for Future AGI evals, and it does not call the API itself.
Install
pip install traceAI-typesafe typesafe-sdk
Use
from fi_instrumentation import register
from fi_instrumentation.fi_types import ProjectType
from traceai_typesafe import TypeSafeInstrumentor
from typesafe_sdk import Choice, Noul, TypeSafeClient
trace_provider = register(project_type=ProjectType.OBSERVE, project_name="support-triage")
TypeSafeInstrumentor().instrument(tracer_provider=trace_provider)
with TypeSafeClient() as client: # reads TYPESAFE_API_KEY
result = client.system_one(
state="I was charged twice. Please help.",
questions={
"billing": Noul(instructions="Is this about billing?"),
"tone": Choice(instructions="What is the tone?", criteria={"calm": None, "angry": None}),
},
)
trace_provider.force_flush()
The instrumentor never sets base_url or a model. Point the client wherever you already do
(base_url= or TYPESAFE_BASE_URL); server.address records the host.
Privacy
Like every traceAI instrumentor, content is captured by default:
| Setting | Effect |
|---|---|
| default | input.value = the state and the questions (instructions, criteria); output.value and typesafe.question.<name>.choice = the decisions |
FI_HIDE_INPUTS=true (or TraceConfig(hide_inputs=True), FI_HIDE_INPUT_TEXT) |
no input.value, no typesafe.idempotency_key |
FI_HIDE_OUTPUTS=true (or TraceConfig(hide_outputs=True), FI_HIDE_OUTPUT_TEXT) |
no output.value, no choice labels; numeric answers stay |
FI_PII_REDACTION=true |
regex PII redaction on every string attribute and the error message |
Always recorded: question names and types (they are attribute keys, so do not put secrets in question names), numeric answers, model, usage, request id. Never recorded: the API key, stack traces, the text of client-side errors (they can name a local file), image bytes or paths. The API's own error message is added to the span status only while neither inputs nor outputs are hidden.
Errors and retries
An HTTP error (401, 403, 422, 429, 5xx) or a connection error ends the span with ERROR,
error.type, http.response.status_code, the server's typesafe.error_type and the request
id. The instrumentor adds no retries. The SDK's own RetryPolicy (by default up to 2 retries
for 408, 429 and 5xx) runs inside the one span, so a 429 followed by a success is one OK
span. A call you repeat yourself is a second span.
Tests
pip install -e . pytest pytest-asyncio
pytest
The tests run the real typesafe-sdk clients against an in-process fake shaped like
TypeSafe's published OpenAPI document. They never contact api.typesafe.ai.
Metadata
Release files for traceAI-typesafe 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 |
|---|---|---|---|---|
| traceai_typesafe-0.1.0.tar.gz | 12.2 kB | Without provenance | Details for traceai_typesafe-0.1.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| traceai_typesafe-0.1.0-py3-none-any.whl | Python 3 | none | any | Details for traceai_typesafe-0.1.0-py3-none-any.whl |
Total release size: 24.6 kB
Release files / traceai_typesafe-0.1.0.tar.gz
| Download URL | traceai_typesafe-0.1.0.tar.gz |
|---|---|
| Size | 12.2 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
3ac1e52bd471881fdbd059fd8677459beb04415aae0ca9884186182e94880223
|
|
BLAKE2b-256 checksum How to use checksums |
bf1976f42f5f04c62cabdafdc168de135b1101a69884bccca8d5b61ec6bddfe6
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Release files / traceai_typesafe-0.1.0-py3-none-any.whl
| Download URL | traceai_typesafe-0.1.0-py3-none-any.whl |
|---|---|
| Size | 12.3 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
3749d0370f057a92e5d4030b33c47e5acb086b53feec26d4b193dc0aff39fc0e
|
|
BLAKE2b-256 checksum How to use checksums |
ee2ee32e34aa60d7c3cd856fa8480d7d81b5f8a2c1e7f5a02d233d9c2962207c
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Security for traceAI-typesafe 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.
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.