vrv-panw · Sep 29, 2026 · Sep 27, 2026 · Sep 28, 2026 · Sep 29, 2026
diff --git a/aigw/product/ai-gateway/decisions-api.mdx b/aigw/product/ai-gateway/decisions-api.mdx
@@ -0,0 +1,90 @@
+---
+title: "Decisions"
+description: "Get typed judgments (probability, choice, or score) from models through the Prisma AIRS AI Gateway's /v1/decisions endpoint."
+---
+
+The Decisions API returns a **typed judgment** instead of generated text. You send a `state` and a set of `questions`. The model returns one typed answer for each question.
+
+Use it for classification, routing, scoring, and yes/no checks, where you need a structured answer and not free text.
+
+<Note>
+  `/v1/decisions` is currently available through [TypeSafe's Jev models](/aigw/integrations/llms/typesafe). It does not support streaming.
+</Note>
+
+## Quick Start
+
+```sh cURL
+curl https://aigw.portkey.ai/v1/decisions \
+  -H "Content-Type: application/json" \
+  -H "Authorization: Bearer $PORTKEY_API_KEY" \
+  -H "x-portkey-provider: typesafe" \
+  -d '{
+    "model": "jev-latest",
+    "state": "Help! My payouts have been failing for 3 days.",
+    "questions": {
+      "department": {
+        "type": "choice",
+        "instructions": "Which team should handle this?",
+        "criteria": { "billing": "Payments", "technical": "Bugs" }
+      }
+    }
+  }'
+```
+
+Response:
+
+```json
+{
+  "model": "jev-1.13.0",
+  "answers": {
+    "department": {
+      "type": "choice",
+      "choice": "billing",
+      "probabilities": { "billing": 0.88, "technical": 0.12 },
+      "confidence": 0.81
+    }
+  },
+  "usage": { "input_tokens": 296, "output_tokens": 20 },
+  "provider": "typesafe"
+}
+```
+
+## Question Types
+
+Each entry in `questions` sets a `type`. The type sets the shape of the answer:
+
+| Type | Use for | Answer shape |
+|:---|:---|:---|
+| `noul` | A yes/no judgment | `{ noul: <0–1 probability> }` |
+| `choice` | One label from a fixed set | `{ choice, probabilities, confidence }` |
+| `score` | A position on a scale | `{ score, legend, probabilities, confidence }` |
+
+`criteria` gives the type-specific detail. For a `choice` question, it is the set of labels.
+
+## Gateway Features
+
+Decisions requests go through the same gateway as other endpoints:
+
+- **Routing** — Retries, fallbacks, load balancing, and circuit breaking through [Configs](/aigw/product/ai-gateway/configs)
+- **Caching** — [Simple and semantic caching](/aigw/product/ai-gateway/cache-simple-and-semantic)
+- **Guardrails** — Before-request checks on the `state` field only. See [Guardrails for Decisions](/aigw/product/guardrails/decisions-guardrails)
+- **Observability** — Logs, cost tracking, and analytics in [Observability](/aigw/product/observability)
+
+## API Reference
+
+- [Create a Decision](/aigw/api-reference/decisions/create-decisions) -- `POST /v1/decisions`
+
+<CardGroup cols={2}>
+  <Card title="TypeSafe" icon="bolt" href="/aigw/integrations/llms/typesafe">
+    Provider setup and pricing
+  </Card>
+  <Card title="API Reference" icon="code" href="/aigw/api-reference/decisions/create-decisions">
+    Decisions API reference
+  </Card>
+  <Card title="Universal API" icon="arrows-rotate" href="/aigw/product/ai-gateway/universal-api">
+    All API formats
+  </Card>
+  <Card title="Guardrails for Decisions" icon="shield" href="/aigw/product/guardrails/decisions-guardrails">
+    Scan `state` before it reaches the model
+  </Card>
+</CardGroup>
diff --git a/aigw/product/ai-gateway/universal-api.mdx b/aigw/product/ai-gateway/universal-api.mdx
@@ -400,7 +400,7 @@ The AI Gateway blocks requests to private and reserved IP ranges by default to p
 - **Gateway to Other APIs** — Proxy requests to any provider endpoint
 - **[Assistants API](/aigw/api-reference/assistants/create-assistant)** — OpenAI Assistants with persistent threads
 - **[Completions](/aigw/api-reference/completions/create-completion)** — Legacy text completion endpoint
-- **Decisions** — Typed judgments (choice, score, probability) via `/v1/decisions`, currently available through [TypeSafe's Jev models](/aigw/integrations/llms/typesafe)
+- **[Decisions](/aigw/product/ai-gateway/decisions-api)** — Typed judgments (choice, score, probability) via `/v1/decisions`, currently available through [TypeSafe's Jev models](/aigw/integrations/llms/typesafe)
 
 ### Multimodal Capabilities
 

diff --git a/api-reference/inference-api/decisions.mdx b/api-reference/inference-api/decisions.mdx
@@ -0,0 +1,4 @@
+---
+title: "Decisions"
+openapi: post /decisions
+---
diff --git a/docs.json b/docs.json
@@ -63,7 +63,8 @@
                           "product/ai-gateway/universal-api",
                           "product/ai-gateway/chat-completions",
                           "product/ai-gateway/messages-api",
-                          "product/ai-gateway/responses-api"
+                          "product/ai-gateway/responses-api",
+                          "product/ai-gateway/decisions-api"
                         ]
                       },
                       "product/ai-gateway/configs",
@@ -808,6 +809,12 @@
                   "api-reference/inference-api/rerank"
                 ]
               },
+              {
+                "group": "Decisions",
+                "pages": [
+                  "api-reference/inference-api/decisions"
+                ]
+              },
               {
                 "group": "OCR",
                 "pages": [
@@ -1486,7 +1493,8 @@
                           "aigw/product/ai-gateway/universal-api",
                           "aigw/product/ai-gateway/chat-completions",
                           "aigw/product/ai-gateway/messages-api",
-                          "aigw/product/ai-gateway/responses-api"
+                          "aigw/product/ai-gateway/responses-api",
+                          "aigw/product/ai-gateway/decisions-api"
                         ]
                       },
                       "aigw/product/ai-gateway/configs",
@@ -2269,6 +2277,16 @@
                       "POST /completions"
                     ]
                   },
+                  {
+                    "group": "Decisions",
+                    "openapi": {
+                      "source": "https://raw.githubusercontent.com/PaloAltoNetworks/openapi/refs/heads/main/openapi.yaml",
+                      "directory": "aigw/api-reference"
+                    },
+                    "pages": [
+                      "POST /decisions"
+                    ]
+                  },
                   {
                     "group": "Embeddings",
                     "openapi": {
@@ -3335,6 +3353,12 @@
                   "api-reference/inference-api/rerank"
                 ]
               },
+              {
+                "group": "Decisions",
+                "pages": [
+                  "api-reference/inference-api/decisions"
+                ]
+              },
               {
                 "group": "OCR",
                 "pages": [

diff --git a/product/ai-gateway/decisions-api.mdx b/product/ai-gateway/decisions-api.mdx
@@ -0,0 +1,100 @@
+---
+title: "Decisions"
+description: "Get typed judgments (probability, choice, or score) from models through Portkey's /v1/decisions endpoint."
+---
+
+<Info>
+Available on all Portkey [plans](https://portkey.ai/pricing).
+</Info>
+
+The Decisions API returns a **typed judgment** instead of generated text. You send a `state` and a set of `questions`. The model returns one typed answer for each question.
+
+Use it for classification, routing, scoring, and yes/no checks, where you need a structured answer and not free text.
+
+<Note>
+  `/v1/decisions` is currently available through [TypeSafe's Jev models](/integrations/llms/typesafe). It does not support streaming.
+</Note>
+
+## Quick Start
+
+```sh cURL
+curl https://api.portkey.ai/v1/decisions \
+  -H "Content-Type: application/json" \
+  -H "x-portkey-api-key: $PORTKEY_API_KEY" \
+  -H "x-portkey-provider: typesafe" \
+  -H "Authorization: Bearer $TYPESAFE_API_KEY" \
+  -d '{
+    "model": "jev-latest",
+    "state": "Help! My payouts have been failing for 3 days.",
+    "questions": {
+      "department": {
+        "type": "choice",
+        "instructions": "Which team should handle this?",
+        "criteria": { "billing": "Payments", "technical": "Bugs" }
+      }
+    }
+  }'
+```
+
+Response:
+
+```json
+{
+  "model": "jev-1.13.0",
+  "answers": {
+    "department": {
+      "type": "choice",
+      "choice": "billing",
+      "probabilities": { "billing": 0.88, "technical": 0.12 },
+      "confidence": 0.81
+    }
+  },
+  "usage": { "input_tokens": 296, "output_tokens": 20 },
+  "provider": "typesafe"
+}
+```
+
+## Question Types
+
+Each entry in `questions` sets a `type`. The type sets the shape of the answer:
+
+| Type | Use for | Answer shape |
+|:---|:---|:---|
+| `noul` | A yes/no judgment | `{ noul: <0–1 probability> }` |
+| `choice` | One label from a fixed set | `{ choice, probabilities, confidence }` |
+| `score` | A position on a scale | `{ score, legend, probabilities, confidence }` |
+
+`criteria` gives the type-specific detail. For a `choice` question, it is the set of labels.
+
+## Gateway Features
+
+Decisions requests go through the same gateway as other endpoints:
+
+- **Routing** — Retries, fallbacks, load balancing, and circuit breaking through [Configs](/product/ai-gateway/configs)
+- **Caching** — [Simple and semantic caching](/product/ai-gateway/cache-simple-and-semantic)
+- **Guardrails** — Before-request checks on the `state` field only. See [Guardrails for Decisions](/product/guardrails/decisions-guardrails)
+- **Observability** — Logs, cost tracking, and analytics in [Observability](/product/observability)
+
+## API Reference
+
+- [Create a Decision](/api-reference/inference-api/decisions) -- `POST /v1/decisions`
+
+<CardGroup cols={2}>
+  <Card title="TypeSafe" icon="bolt" href="/integrations/llms/typesafe">
+    Provider setup and pricing
+  </Card>
+  <Card title="API Reference" icon="code" href="/api-reference/inference-api/decisions">
+    Decisions API reference
+  </Card>
+  <Card title="Universal API" icon="arrows-rotate" href="/product/ai-gateway/universal-api">
+    All API formats
+  </Card>
+  <Card title="Guardrails for Decisions" icon="shield" href="/product/guardrails/decisions-guardrails">
+    Scan `state` before it reaches the model
+  </Card>
+</CardGroup>
+
+
+import PrismaAirsCta from "/snippets/prisma-airs-cta.mdx";
+
+<PrismaAirsCta />
diff --git a/product/ai-gateway/universal-api.mdx b/product/ai-gateway/universal-api.mdx
@@ -552,7 +552,7 @@ Portkey blocks requests to private and reserved IP ranges by default to prevent
 - **[Gateway to Other APIs](/api-reference/inference-api/gateway-for-other-apis)** — Proxy requests to any provider endpoint
 - **[Assistants API](/api-reference/inference-api/assistants-api/assistants/create-assistant)** — OpenAI Assistants with persistent threads
 - **[Completions](/api-reference/inference-api/completions)** — Legacy text completion endpoint
-- **Decisions** — Typed judgments (choice, score, probability) via `/v1/decisions`, currently available through [TypeSafe's Jev models](/integrations/llms/typesafe)
+- **[Decisions](/product/ai-gateway/decisions-api)** — Typed judgments (choice, score, probability) via `/v1/decisions`, currently available through [TypeSafe's Jev models](/integrations/llms/typesafe)
 
 ### Multimodal Capabilities