golem-http-params-ts

HTTP Request and Response Parameter Mapping (TypeScript)

Overview

When an agent is exposed over HTTP, Golem maps parts of each HTTP request to the agent id record and to method inputs. This skill covers how path segments, query parameters, headers, and request bodies are mapped, which schema types are supported for each, and how return schemas map to HTTP responses.

Path Variables

Path variables {varName} map by name — mount-path variables bind to the agent's id fields, endpoint-path variables bind to the method's input keys:

import { z } from 'zod';
import { defineAgent, method, http } from '@golemcloud/golem-ts-sdk';

export const TaskAgent = defineAgent({
  name: 'TaskAgent',
  id: { name: z.string() },                 // {name} in the mount path
  http: http.mount('/api/tasks/{name}'),
  methods: {
    // {itemId} in the endpoint path binds to the `itemId` input
    getItem: method({ input: { itemId: z.string() }, returns: Item, http: http.get('/items/{itemId}') }),
  },
});

Remaining (catch-all) path variables capture everything after a prefix:

getFile: method({ input: { path: z.string() }, returns: FileContent, http: http.get('/files/{*path}') }),
// GET .../files/docs/readme.md → path = "docs/readme.md"

Catch-all variables can only appear as the last endpoint path segment and are not allowed in mount paths.

Query Parameters

Bind query parameters either inline in the endpoint path with ?key={var} syntax, or explicitly via the query option map (its values are input keys):

// Inline form
search: method({
  input: { query: z.string(), maxResults: z.number() },
  returns: z.array(SearchResult),
  http: http.get('/search?q={query}&limit={maxResults}'),
}),

// Explicit map form (query-param name → input key)
search2: method({
  input: { query: z.string(), maxResults: z.number() },
  returns: z.array(SearchResult),
  http: http.get('/search', { query: { q: 'query', limit: 'maxResults' } }),
}),

Header Variables

Bind HTTP headers to inputs with the headers option map (header name → input key):

getData: method({
  input: { requestId: z.string(), token: z.string() },
  returns: Data,
  http: http.get('/data', { headers: { 'X-Request-Id': 'requestId', 'Authorization': 'token' } }),
}),

Note: Unlike the old decorator API, the fluent http.mount(...) does not support a mount-level headers map. Every agent id field must be supplied by a mount path variable. Header binding is available on endpoints only, and binds to method inputs.

Supported Schema Types for Path, Query, and Header Variables

Only scalar schemas can be used for inputs bound to path/query/header variables (the value is parsed from the URL/header string):

For query parameters and headers only (not path variables), two additional wrapper forms are supported:

All other schemas (z.object(...), nested arrays, maps, etc.) can only be used as body parameters.

POST Request Body Mapping

For POST/PUT/DELETE/PATCH endpoints, any method input not bound to a path variable, query parameter, or header is populated from the JSON request body:

updateItem: method({
  input: { id: z.string(), name: z.string(), count: z.number() },
  returns: Item,
  http: http.post('/items/{id}'),
}),
// POST .../items/123
// Body: { "name": "Widget", "count": 5 }
// → id from path, name and count from body

Each unbound input becomes a top-level field in the expected JSON body object. Field names use the input key names.

⚠️ Important: The request body is always a JSON object with input keys as fields — even when there is only a single body parameter. For example, a method with input: { decision: z.string() } expects {"decision": "approved"}, never a bare string like "approved". Sending a non-object JSON value or plain text will fail with REQUEST_JSON_BODY_PARSING_FAILED.

Binary Request and Response Bodies

Use the s.unstructuredBinary() schema marker for raw binary payloads. Its decoded value is a reference: { tag: 'url', val: string } or { tag: 'inline', val: Uint8Array, mimeType?: string }. A method using it may have only one body parameter, and that input cannot also be bound to a path/query/header variable.

import { z } from 'zod';
import { defineAgent, method, http, s } from '@golemcloud/golem-ts-sdk';

export const UploadAgent = defineAgent({
  name: 'UploadAgent',
  id: { name: z.string() },
  http: http.mount('/uploads/{name}'),
  methods: {
    // Accepting any binary content type
    upload: method({
      input: { bucket: z.string(), payload: s.unstructuredBinary() },
      returns: z.number(),
      http: http.post('/upload/{bucket}'),
    }),
    // Restricting to specific MIME types
    uploadImage: method({
      input: { bucket: z.string(), payload: s.unstructuredBinary({ mimeTypes: ['image/gif'] }) },
      returns: z.number(),
      http: http.post('/upload-image/{bucket}'),
    }),
    // Returning binary data
    download: method({ input: {}, returns: s.unstructuredBinary(), http: http.get('/download') }),
  },
});

export const UploadAgentImpl = UploadAgent.implement({
  init: () => ({}),
  methods: {
    upload({ payload }) {
      if (payload.tag === 'url') return -1;
      return payload.val.byteLength;
    },
    uploadImage({ payload }) {
      return payload.tag === 'inline' ? payload.val.byteLength : -1;
    },
    download() {
      return { tag: 'inline', val: new Uint8Array([1, 2, 3]), mimeType: 'application/octet-stream' };
    },
  },
});

Plain Text Request and Response Bodies

Use the s.unstructuredText() marker for raw text/plain payloads. Its decoded value is { tag: 'url', val: string } or { tag: 'inline', val: string, languageCode?: string }. Like binary, a method using it may have only one body parameter, not bound to any path/query/header variable. The body is decoded as UTF-8.

methods: {
  // Accepting any text/plain content
  addNote: method({
    input: { id: z.string(), body: s.unstructuredText() },
    returns: z.number(),
    http: http.post('/notes/{id}'),
  }),
  // Restricting to specific language codes
  translate: method({
    input: { id: z.string(), body: s.unstructuredText({ languages: ['en', 'de'] }) },
    returns: z.string(),
    http: http.post('/translate/{id}'),
  }),
  // Returning text/plain
  getNote: method({ input: { id: z.string() }, returns: s.unstructuredText(), http: http.get('/notes/{id}') }),
}

// Handler returning inline text with a language code:
getNote({ id }) {
  return { tag: 'inline', val: 'hello', languageCode: 'en' };
}

HTTP-level rules:

• The request must have either no Content-Type, text/plain, or text/plain; charset=utf-8 (case-insensitive). Any other content type is rejected with 415 Unsupported Media Type.
• Content-Language is always optional, even when language codes are restricted. If present, it must be a single value (multi-valued or comma-separated headers are rejected with 400 Bad Request).
• When restricted, the supplied Content-Language is matched case-insensitively against the allowed list; otherwise 415 Unsupported Media Type.
• A non-UTF-8 request body is rejected with 400 Bad Request.
• Content-Language cannot also be bound as an endpoint header parameter when the body is unstructured text — that header is reserved for declaring the body language.

The response is sent as Content-Type: text/plain; charset=utf-8. If the returned inline text has a languageCode, it is forwarded as the Content-Language response header.

Return Schema to HTTP Response Mapping

Data Type to JSON Mapping

Accessing the Principal

When authentication is enabled, access the authenticated user via this.getPrincipal() inside the handler (it is not a method input). See the golem-add-http-auth-ts skill:

whoAmI() {
  const principal = this.getPrincipal();
  return { value: principal };
}

Related Skills

• Load golem-add-http-endpoint-ts for the high-level workflow of defining and mounting HTTP endpoints on an agent