Usage>API Reference

API Reference

All endpoints exposed by the Prism proxy server. By default the server binds to :8080.

Base URL

text

http://localhost:8080

Common request headers

Header	Required	Description
Authorization	Yes	Your LLM provider API key in Bearer sk-... format. Forwarded to the upstream provider on cache miss.
Content-Type	Yes	Must be application/json.
X-Tenant-ID	No	Namespace for cache isolation. Defaults to default if omitted. See Multi-Tenant.

Response headers

Header	Values	Description
X-Cache-Hit	true / false	Whether the response was served from cache.
X-Cache	L1 / L2a / L2b / backend	Which layer served the request.

Endpoints

POST/proxy/openaiOpenAI Chat Completions proxy

Forwards to api.openai.com/v1/chat/completions. Accepts the full OpenAI request schema including streaming, function calling, and vision.

POST/proxy/anthropicAnthropic Messages API proxy

Forwards to api.anthropic.com/v1/messages. Supports all Claude models and tool use.

POST/proxy/groqGroq Chat Completions proxy

Forwards to api.groq.com/openai/v1/chat/completions. Uses the OpenAI-compatible schema.

POST/proxy/togetherTogether AI proxy

Forwards to api.together.xyz/v1/chat/completions. Uses the OpenAI-compatible schema.

GET/healthHealth check

Returns {"status":"ready"} when the proxy is fully initialised (DB migrations run, embedding model loaded).

GET/metricsPrometheus metrics

Exposes cache_hits_total, cache_misses_total, request_latency_histogram, and upstream_errors_total labelled by layer and tenant.

GET/analytics/cost-savingsCost savings report

Returns per-tenant saved token counts and estimated USD savings since last reset.

POST/cache/queryDirect cache query

Legacy endpoint for querying the cache directly without proxying to an upstream API. Useful for tooling and debugging.

Health check example

bash

curl http://localhost:8080/health
# → {"status":"ready"}

Cost savings response

json

GET /analytics/cost-savings

{
  "tenants": [
    {
      "id": "team-alpha",
      "cached_requests": 14823,
      "tokens_saved": 2940000,
      "estimated_usd_saved": 88.20
    },
    {
      "id": "default",
      "cached_requests": 3201,
      "tokens_saved": 640200,
      "estimated_usd_saved": 19.21
    }
  ],
  "total_usd_saved": 107.41
}

Prometheus metrics

bash

# Fetch all metrics
curl http://localhost:8080/metrics

# Filter for cache stats
curl -s http://localhost:8080/metrics | grep cache_hits
# cache_hits_total{layer="L1",tenant="team-alpha"} 9241
# cache_hits_total{layer="L2a",tenant="team-alpha"} 3991
# cache_hits_total{layer="L2b",tenant="team-alpha"} 1591

PreviousIntegration NextConfiguration