ai() API
LLM provider, model, media, routing, and embedding APIs.
Most Used
ai
Create a provider client from a named deployment profile and options.
| Field | Value |
|---|---|
| Kind | function |
| Ax concept | ai |
| API section | AxAI |
| Form | ai(provider='openai', **options) |
| Returns | AI client/service |
Important Options
api key or credential providermodelapi URLheaderstransport
Examples
client = ai("openai", api_key=os.environ["OPENAI_API_KEY"])OpenAICompatibleClient
OpenAI-compatible chat, stream, embedding, audio, and realtime provider boundary.
| Field | Value |
|---|---|
| Kind | type |
| Ax concept | OpenAICompatibleClient |
| API section | AxAI |
| Form | OpenAICompatibleClient(options=None) |
| Returns | provider client |
Important Options
api keymodelbase URLtransport
OpenAIResponsesClient
OpenAI Responses provider mapping using the same Core-owned request and response contract.
| Field | Value |
|---|---|
| Kind | type |
| Ax concept | OpenAIResponsesClient |
| API section | AxAI |
| Form | OpenAIResponsesClient(options=None) |
| Returns | provider client |
Important Options
api keymodelaudiorealtime
GoogleGeminiClient
Gemini provider mapping for chat, streaming, media, tools, embeddings, and usage normalization.
| Field | Value |
|---|---|
| Kind | type |
| Ax concept | GoogleGeminiClient |
| API section | AxAI |
| Form | GoogleGeminiClient(options=None) |
| Returns | provider client |
Important Options
api keymodelembed model
AnthropicClient
Anthropic provider mapping for messages, thinking, cache control, streaming, and usage normalization.
| Field | Value |
|---|---|
| Kind | type |
| Ax concept | AnthropicClient |
| API section | AxAI |
| Form | AnthropicClient(options=None) |
| Returns | provider client |
Important Options
api keymodelthinkingcache control
AxBalancer
Retry and route requests across multiple provider services, with opt-in adaptive cost, reliability, and deadline routing.
| Field | Value |
|---|---|
| Kind | type |
| Ax concept | AxBalancer |
| API section | AxAI |
| Form | AxBalancer(services, options=None) |
| Returns | AI service |
Important Options
servicesretry policycapability requirementsadaptive strategy
MultiServiceRouter
Choose a service by capability or model routing policy.
| Field | Value |
|---|---|
| Kind | type |
| Ax concept | MultiServiceRouter |
| API section | AxAI |
| Form | MultiServiceRouter(services) |
| Returns | AI service |
Important Options
servicesrouting
ProviderRouter
Route requests by capability, preserving native media for supporting providers and degrading only when required.
| Field | Value |
|---|---|
| Kind | type |
| Ax concept | ProviderRouter |
| API section | AxAI |
| Form | ProviderRouter(providers, routing=None, processing=None) |
| Returns | AI service |
Important Options
providersroutingprocessing
Advanced / internals / full reference
get_supported_ai_models
Return the AxIR-backed provider and model catalog, including dynamic named profiles and portable capability metadata.
| Field | Value |
|---|---|
| Kind | function |
| Ax concept | get_supported_ai_models |
| API section | AxAI |
| Form | get_supported_ai_models(model_type=None) |
| Returns | provider catalog entries |
Important Options
type filterthinkingLevelsserviceTiersisDynamic
Examples
catalog = get_supported_ai_models()dict[str, str]
Request metadata passed to a renewable credential callback for every transport attempt.
| Field | Value |
|---|---|
| Kind | type |
| Ax concept | AxCredentialRequest |
| API section | AxAI |
| Form | dict[str, str] |
| Returns | credential request |
Important Options
profileoperationmethodURL
Callable[[dict[str, str]], dict[str, str]]
Return fresh request headers that override static profile authentication.
| Field | Value |
|---|---|
| Kind | interface |
| Ax concept | AxCredentialProvider |
| API section | AxAI |
| Form | Callable[[dict[str, str]], dict[str, str]] |
| Returns | header map or credential error |
Important Options
chatstreamembeddingsResponsesaudioretries
closable generator
Incremental, closeable provider event stream. Retry and failover stop once content is delivered.
| Field | Value |
|---|---|
| Kind | type |
| Ax concept | AxProviderStream |
| API section | AxAI |
| Form | client.stream(request) |
| Returns | incremental chat events |
Important Options
next eventterminal errorconsumer cancellationupstream close
Examples
for event in client.stream(request): handle(event)AxUsageContext
Application attribution merged from service defaults and per-call overrides.
| Field | Value |
|---|---|
| Kind | type |
| Ax concept | AxUsageContext |
| API section | AxAI |
| Form | dict[str, object] |
| Returns | usage context |
Important Options
tenantuserrequestrunfeatureattributes
AxUsageEvent
Normalized token usage and correlation data for one completed chat or embedding operation.
| Field | Value |
|---|---|
| Kind | type |
| Ax concept | AxUsageEvent |
| API section | AxAI |
| Form | AxUsageEvent |
| Returns | usage event |
Important Options
providermodeltokenscontextcorrelation IDsstreaming
AxUsageObserver
Best-effort process-wide callback for normalized usage events.
| Field | Value |
|---|---|
| Kind | interface |
| Ax concept | AxUsageObserver |
| API section | AxAI |
| Form | AxUsageObserver |
| Returns | usage observer |
Important Options
fail-open deliverysynchronous enqueue
set_usage_observer
Register, replace, or clear the process-wide usage observer.
| Field | Value |
|---|---|
| Kind | function |
| Ax concept | set_usage_observer |
| API section | AxAI |
| Form | set_usage_observer(observer) |
| Returns | void |
Important Options
observerclear
Examples
set_usage_observer(events.append)AxRuntimeHooks
Non-serializable rate limiter, tracer, and meter overrides for one service or program execution.
| Field | Value |
|---|---|
| Kind | type |
| Ax concept | AxRuntimeHooks |
| API section | AxAI |
| Form | AxRuntimeHooks(rate_limiter=None, tracer=None, meter=None) |
| Returns | runtime hooks |
Important Options
rate limitertracermeterprecedence
Examples
hooks = AxRuntimeHooks(rate_limiter=limiter, tracer=tracer, meter=meter)AxRateLimitInfo
Provider and model metadata supplied to a request-wrapper rate limiter.
| Field | Value |
|---|---|
| Kind | type |
| Ax concept | AxRateLimitInfo |
| API section | AxAI |
| Form | AxRateLimitInfo |
| Returns | rate-limit context |
Important Options
operationprovidermodelstreamingprevious usage
AxRateLimiter
Request wrapper that can delay, reject, or execute a chat or embedding request.
| Field | Value |
|---|---|
| Kind | interface |
| Ax concept | AxRateLimiter |
| API section | AxAI |
| Form | AxRateLimiter |
| Returns | provider result |
Important Options
nextmetadataerrors
AxTracer
Dependency-neutral tracer that creates metadata-only Ax spans.
| Field | Value |
|---|---|
| Kind | interface |
| Ax concept | AxTracer |
| API section | AxAI |
| Form | AxTracer |
| Returns | span |
Important Options
start spanparent spanfail-open
AxMeter
Dependency-neutral meter for Ax counters, histograms, and gauges.
| Field | Value |
|---|---|
| Kind | interface |
| Ax concept | AxMeter |
| API section | AxAI |
| Form | AxMeter |
| Returns | instrument |
Important Options
counterhistogramgaugefail-open
AxGlobals
Process-wide live defaults for runtime hooks and usage observation.
| Field | Value |
|---|---|
| Kind | type |
| Ax concept | AxGlobals |
| API section | AxAI |
| Form | AxGlobals |
| Returns | global hooks |
Important Options
rate limitertracermeterusage observer
set_rate_limiter
Register, replace, or clear the process-wide request-wrapper rate limiter.
| Field | Value |
|---|---|
| Kind | function |
| Ax concept | set_rate_limiter |
| API section | AxAI |
| Form | set_rate_limiter(limiter) |
| Returns | void |
Important Options
limiterclear
Examples
set_rate_limiter(limiter)set_tracer
Register, replace, or clear the process-wide dependency-neutral tracer.
| Field | Value |
|---|---|
| Kind | function |
| Ax concept | set_tracer |
| API section | AxAI |
| Form | set_tracer(tracer) |
| Returns | void |
Important Options
tracerclear
Examples
set_tracer(tracer)set_meter
Register, replace, or clear the process-wide dependency-neutral meter.
| Field | Value |
|---|---|
| Kind | function |
| Ax concept | set_meter |
| API section | AxAI |
| Form | set_meter(meter) |
| Returns | void |
Important Options
meterclear
Examples
set_meter(meter)AxBalancerAdaptiveStrategy
Configure adaptive provider routing without changing the ordered default.
| Field | Value |
|---|---|
| Kind | type |
| Ax concept | AxBalancerAdaptiveStrategy |
| API section | AxAI |
| Form | AxBalancerAdaptiveStrategy |
| Returns | adaptive strategy |
Important Options
deadlinebad outcome costexpected tokensstable route keysslicestats storerouting events
AxBalancerStatsStore
Store shared adaptive decision state with atomic observations.
| Field | Value |
|---|---|
| Kind | interface |
| Ax concept | AxBalancerStatsStore |
| API section | AxAI |
| Form | AxBalancerStatsStore |
| Returns | stats store |
Important Options
getobserve
AxInMemoryBalancerStatsStore
Thread-safe in-memory adaptive stats store.
| Field | Value |
|---|---|
| Kind | type |
| Ax concept | AxInMemoryBalancerStatsStore |
| API section | AxAI |
| Form | AxInMemoryBalancerStatsStore |
| Returns | stats store |
create_balancer_route_stats
Create neutral adaptive route statistics.
| Field | Value |
|---|---|
| Kind | function |
| Ax concept | create_balancer_route_stats |
| API section | AxAI |
| Form | create_balancer_route_stats |
| Returns | route stats |
update_balancer_route_stats
Purely reduce one success or failure observation into route statistics.
| Field | Value |
|---|---|
| Kind | function |
| Ax concept | update_balancer_route_stats |
| API section | AxAI |
| Form | update_balancer_route_stats |
| Returns | route stats |
Important Options
current statsobservation
sample_balancer_route_health
Sample failure and deadline-miss probability for adaptive exploration.
| Field | Value |
|---|---|
| Kind | function |
| Ax concept | sample_balancer_route_health |
| API section | AxAI |
| Form | sample_balancer_route_health |
| Returns | sampled health |
Important Options
route statsdeadline
Full Reference
Open the full Python reference. This page highlights the ai symbols most relevant to everyday use.