merge main

Author: H2Shami

.husky/pre-push

@@ -0,0 +1,11 @@
+#!/bin/bash
+protected_branch='main'
+current_branch=$(git symbolic-ref HEAD | sed -e 's,.*/\(.*\),\1,')
+
+if [ "$current_branch" = "$protected_branch" ]; then
+    echo "🚫 Direct push to main blocked!"
+    echo "Create a feature branch and open a PR instead."
+    exit 1
+fi
+
+exit 0

CLAUDE.md

@@ -169,6 +169,18 @@ className="bg-background text-foreground border-border"
 - Test in both light and dark modes
 - when making changes to /packages run the tests npx jest **tests**/ in /packages to make sure nothing else is broken
 
+### TypeScript Type Changes
+
+When modifying TypeScript types (interfaces, type aliases, union types):
+
+1. **Run type checking before committing**: `npx tsc --noEmit` in the relevant package to catch type errors (Jest tests alone won't catch all type mismatches)
+2. **Search for related types**: When updating union types (e.g., `"low" | "medium" | "high"`), grep for similar patterns to ensure all related types are updated:
+   ```bash
+   grep -r "thinkingLevel" packages/  # Find all usages of similar type
+   ```
+3. **Check cross-package dependencies**: Types in `/packages/` are often used across multiple packages (llm-mapper, prompts, cost). Verify changes don't break consumers.
+4. **Run the full build**: For `/packages/` changes, run `cd web && yarn build` or check Vercel preview to catch type errors across the monorepo.
+
 # Helicone Design System Guidelines
 
 ## Core Principles

docs/docs.json

@@ -324,6 +324,7 @@
               "rest/request/post-v1requestquery-clickhouse",
               "rest/request/post-v1requestquery",
               "rest/request/get-v1request",
+              "rest/request/get-v1request-inputs",
               "rest/request/post-v1requestquery-ids",
               "rest/request/post-v1request-feedback",
               "rest/request/put-v1request-property",

docs/rest/request/get-v1request-inputs.mdx

@@ -0,0 +1,53 @@
+---
+title: "Get Request Inputs"
+sidebarTitle: "Get Request Inputs"
+description: "Retrieve the prompt template inputs (variables) used for a specific request made through AI Gateway prompt management."
+"twitter:title": "Get Request Inputs - Helicone OSS LLM Observability"
+openapi: get /v1/request/{requestId}/inputs
+---
+
+import EUAPIWarning from "/snippets/eu-api-warning.mdx";
+
+<EUAPIWarning />
+
+## Overview
+
+When you use [Prompt Management](/features/advanced-usage/prompts/overview) through the AI Gateway, template variables (inputs) are stored automatically. This endpoint lets you retrieve those inputs by request ID — useful for building testing pipelines that replay past requests against new prompt versions.
+
+## Use Cases
+
+- **Regression testing**: Pull a past request's inputs and replay them against a new prompt version to validate behavior.
+- **Prompt comparison**: Compare outputs across prompt versions using the same inputs, without storing inputs separately on your end.
+- **Debugging**: Inspect the exact variables that were injected into a prompt template at runtime.
+
+<Note>Request data is retained for 90 days. Plan your testing workflows accordingly.</Note>
+
+## Response
+
+Returns `null` for `data` if the request has no associated inputs (e.g., the request was not made through prompt management, or the request ID doesn't exist).
+
+### Example Response
+
+```json
+{
+  "data": {
+    "inputs": {
+      "customer_name": "Sarah",
+      "issue": "refund request"
+    },
+    "prompt_id": "customer-support",
+    "version_id": "1c7a86c8-...",
+    "environment": "production"
+  },
+  "error": null
+}
+```
+
+### No Inputs Found
+
+```json
+{
+  "data": null,
+  "error": null
+}
+```