<div align="center">
<img src="https://raw.githubusercontent.com/TrustBound/dream/main/ricky_and_lucy.png" alt="Dream Logo" width="200">
<a href="https://hexdocs.pm/dream_http_client">
<img src="https://img.shields.io/badge/hex-docs-lightgreen.svg" alt="HexDocs">
</a>
</div>
# dream_http_client
**Type-safe HTTP client for Gleam with streaming support.**
A standalone HTTP/HTTPS client built on Erlang's battle-tested `httpc`. Supports both streaming and non-streaming requests. Built with the same quality standards as [Dream](https://github.com/TrustBound/dream), but completely independent—use it in any Gleam project.
## Features
- ✅ **Three execution modes**: Blocking, yielder streaming, and message-based streaming
- ✅ **OTP-compatible**: Message-based streams work with actors and selectors
- ✅ **Concurrent streams**: Handle multiple HTTP streams in a single actor
- ✅ **Stream cancellation**: Cancel in-flight requests cleanly
- ✅ **HTTPS support**: Full TLS/SSL support via Erlang
- ✅ **Builder pattern**: Consistent, composable request configuration
- ✅ **Type-safe**: `Result` types force error handling
- ✅ **Battle-tested**: Erlang's `httpc` under the hood
- ✅ **Framework-independent**: Zero dependencies on Dream or other frameworks
## Installation
```bash
gleam add dream_http_client
```
## Quick Start
Choose the execution mode that fits your use case:
### 1. Blocking Requests - `client.send()`
Perfect for JSON APIs and small responses:
```gleam
import dream_http_client/client.{method, scheme, host, path, add_header, send}
import gleam/http.{Get, Https}
let result = client.new
|> method(Get)
|> scheme(Https)
|> host("api.example.com")
|> path("/users/123")
|> add_header("Authorization", "Bearer " <> token)
|> send()
case result {
Ok(body) -> decode_json(body)
Error(msg) -> handle_error(msg)
}
```
### 2. Yielder-Based Streaming - `client.stream_yielder()`
Perfect for AI responses or simple sequential streaming:
```gleam
import dream_http_client/client.{host, path, stream_yielder}
import gleam/yielder.{each}
import gleam/bytes_tree.{to_string}
import gleam/io.{print}
client.new
|> host("api.openai.com")
|> path("/v1/chat/completions")
|> stream_yielder()
|> each(fn(chunk) {
// Process each chunk as it arrives
print(to_string(chunk))
})
```
**Note:** This is a pull-based synchronous API. It blocks while waiting for chunks,
so it's not suitable for OTP actors that need to handle multiple concurrent operations.
### 3. Message-Based Streaming - `client.stream_messages()`
Perfect for OTP actors handling multiple concurrent streams:
```gleam
import dream_http_client/client.{
type RequestId, type StreamMessage, StreamStart, Chunk, StreamEnd, StreamError,
host, path, stream_messages, select_stream_messages, cancel_stream
}
import gleam/otp/actor.{continue}
import gleam/erlang/process.{type Selector, new_selector}
import gleam/dict.{type Dict}
pub type Message {
HttpStream(StreamMessage)
OtherMessage(String)
}
pub type State {
State(active_streams: Dict(RequestId, StreamState))
}
fn handle_message(msg: Message, state: State) {
case msg {
HttpStream(stream_msg) -> {
case stream_msg {
StreamStart(req_id, headers) -> {
// Stream started, headers received
let new_state = track_stream(state, req_id, headers)
continue(new_state)
}
Chunk(req_id, data) -> {
// Data chunk received, process it
let new_state = process_chunk(state, req_id, data)
continue(new_state)
}
StreamEnd(req_id, trailing_headers) -> {
// Stream completed successfully
let new_state = cleanup_stream(state, req_id)
continue(new_state)
}
StreamError(req_id, reason) -> {
// Stream failed, handle error
let new_state = handle_stream_error(state, req_id, reason)
continue(new_state)
}
}
}
OtherMessage(content) -> {
// Handle other actor messages
continue(state)
}
}
}
fn init_selector() -> Selector(Message) {
new_selector()
|> select_stream_messages(HttpStream)
// Can add more selectors for other message types
}
// Start multiple concurrent streams
pub fn start_streams() {
let assert Ok(req_id_1) = client.new
|> host("api.example.com")
|> path("/stream/1")
|> stream_messages()
let assert Ok(req_id_2) = client.new
|> host("api.example.com")
|> path("/stream/2")
|> stream_messages()
// Both streams send messages to your actor concurrently
// RequestId discriminates between them
}
// Cancel a stream if needed
pub fn cancel_if_needed(req_id: RequestId) {
cancel_stream(req_id)
// No more messages will arrive for this stream
}
```
## Choosing an Execution Mode
| Use Case | Recommended Mode | Why |
|----------|------------------|-----|
| JSON API calls | `send()` | Simple, complete response at once |
| Small file downloads | `send()` | Load entire file into memory |
| AI/LLM streaming (single request) | `stream_yielder()` | Sequential token processing |
| Simple file downloads | `stream_yielder()` | Memory-efficient chunked processing |
| OTP actors with multiple streams | `stream_messages()` | Non-blocking, concurrent, cancellable |
| Long-lived connections | `stream_messages()` | Can cancel mid-stream |
| Integration with supervisors | `stream_messages()` | Full OTP compatibility |
## Usage
### Building Requests
All three execution modes use the same builder pattern:
```gleam
import dream_http_client/client.{
method, scheme, host, port, path, query, add_header, body, timeout
}
import gleam/http.{Post, Https}
let request = client.new
|> method(Post)
|> scheme(Https)
|> host("api.example.com")
|> port(443)
|> path("/api/users")
|> query("page=1&limit=10")
|> add_header("Content-Type", "application/json")
|> add_header("Authorization", "Bearer " <> token)
|> body(json_body)
|> timeout(60_000) // 60 second timeout for slow APIs
```
### Blocking Requests - `send()`
Get the complete response body at once:
```gleam
import dream_http_client/client.{send}
import gleam/json.{decode}
case send(request) {
Ok(body) -> {
// Process complete response
case decode(body, user_decoder) {
Ok(user) -> Ok(user)
Error(_) -> Error("Invalid JSON")
}
}
Error(msg) -> Error("Request failed: " <> msg)
}
```
### Yielder Streaming - `stream_yielder()`
Process chunks sequentially as they arrive:
```gleam
import dream_http_client/client.{stream_yielder}
import gleam/yielder.{each, to_list}
import gleam/bytes_tree.{to_string}
import gleam/io.{println}
// Process chunks incrementally
stream_yielder(request)
|> each(fn(chunk) {
let text = to_string(chunk)
println("Received: " <> text)
})
// Or collect all chunks (stream completes automatically)
let chunks = stream_yielder(request)
|> to_list() // No need to use take() - stream stops when done!
```
**Important:** This blocks the calling process while waiting for chunks.
Don't use this in OTP actors that need to handle other messages concurrently.
### Message-Based Streaming - `stream_messages()`
Handle streams in OTP actors with full concurrency:
```gleam
import dream_http_client/client.{stream_messages, select_stream_messages, cancel_stream}
import gleam/erlang/process.{new_selector}
// 1. Start a stream - returns immediately with RequestId
let assert Ok(req_id) = stream_messages(request)
// 2. Messages arrive in your process mailbox automatically:
// - StreamStart(req_id, headers)
// - Chunk(req_id, data) (zero or more)
// - StreamEnd(req_id, headers) or StreamError(req_id, reason)
// 3. Use select_stream_messages to integrate with your selector
let selector = new_selector()
|> select_stream_messages(MyHttpMessage)
// 4. Cancel if needed
cancel_stream(req_id)
```
#### Complete OTP Actor Example
```gleam
import dream_http_client/client.{
type StreamMessage, StreamStart, Chunk, StreamEnd, StreamError,
host, stream_messages, select_stream_messages
}
import gleam/otp/actor.{type Next, type Spec, Spec, Ready, Stop, continue, start_spec}
import gleam/erlang/process.{type Selector, Normal, new_selector, selecting}
import gleam/dict.{type Dict, new as new_dict, insert, get, delete}
import gleam/list.{reverse}
import gleam/bit_array.{concat}
import gleam/string.{inspect}
import gleam/io.{println_error}
pub type Message {
StartDownload(url: String)
HttpStream(StreamMessage)
Stop
}
pub type StreamState {
StreamState(chunks: List(BitArray))
}
pub type State {
State(streams: Dict(String, StreamState))
}
pub fn start() {
start_spec(Spec(
init: fn() {
let selector = new_selector()
|> select_stream_messages(HttpStream)
|> selecting(some_subject, fn(msg) { msg })
Ready(State(streams: new_dict()), selector)
},
init_timeout: 1000,
loop: handle_message,
))
}
fn handle_message(msg: Message, state: State) -> Next(Message, State) {
case msg {
StartDownload(url) -> {
// Start a new stream
let assert Ok(req_id) = client.new
|> host(url)
|> stream_messages()
continue(state)
}
HttpStream(StreamStart(req_id, headers)) -> {
// Stream started, initialize state
let req_id_str = inspect(req_id)
let new_streams = insert(
state.streams,
req_id_str,
StreamState(chunks: [])
)
continue(State(streams: new_streams))
}
HttpStream(Chunk(req_id, data)) -> {
// Accumulate chunk
let req_id_str = inspect(req_id)
case get(state.streams, req_id_str) {
Ok(stream_state) -> {
let updated = StreamState(
chunks: [data, ..stream_state.chunks]
)
let new_streams = insert(state.streams, req_id_str, updated)
continue(State(streams: new_streams))
}
Error(_) -> continue(state)
}
}
HttpStream(StreamEnd(req_id, _headers)) -> {
// Stream complete, process all chunks
let req_id_str = inspect(req_id)
case get(state.streams, req_id_str) {
Ok(stream_state) -> {
// Combine all chunks and process
let complete_data = stream_state.chunks
|> reverse()
|> concat()
process_complete_download(complete_data)
// Remove from active streams
let new_streams = delete(state.streams, req_id_str)
continue(State(streams: new_streams))
}
Error(_) -> continue(state)
}
}
HttpStream(StreamError(req_id, reason)) -> {
// Handle error
println_error("Stream error: " <> reason)
let req_id_str = inspect(req_id)
let new_streams = delete(state.streams, req_id_str)
continue(State(streams: new_streams))
}
Stop -> Stop(Normal)
}
}
```
### POST Requests with JSON
```gleam
import dream_http_client/client.{method, host, path, add_header, body, send}
import gleam/http.{Post}
import gleam/json.{object, string, to_string}
let user_json = object([
#("name", string("Alice")),
#("email", string("alice@example.com")),
])
let result = client.new
|> method(Post)
|> host("api.example.com")
|> path("/users")
|> add_header("Content-Type", "application/json")
|> body(to_string(user_json))
|> send()
```
### Configuring Timeouts
All request types support timeout configuration. Default is 30 seconds:
```gleam
import dream_http_client/client.{host, path, timeout, send}
// Short timeout for quick APIs
let result = client.new
|> host("api.example.com")
|> path("/health")
|> timeout(5_000) // 5 seconds
|> send()
// Long timeout for slow operations
let result = client.new
|> host("ml-api.example.com")
|> path("/train-model")
|> timeout(300_000) // 5 minutes
|> stream_yielder() // Works with all execution modes
```
## API Reference
### Types
#### `ClientRequest`
HTTP request configuration with all components:
- `method`: HTTP method (GET, POST, etc.)
- `scheme`: Protocol (HTTP or HTTPS)
- `host`: Server hostname
- `port`: Optional port number (defaults to 80 for HTTP, 443 for HTTPS)
- `path`: Request path
- `query`: Optional query string
- `headers`: List of header name-value pairs
- `body`: Request body as string
- `timeout`: Optional timeout in milliseconds (defaults to 30000ms)
#### `RequestId`
Opaque identifier for an active message-based stream. Returned from `stream_messages()` and included in all `StreamMessage` variants. Use this to:
- Track multiple concurrent streams
- Cancel specific streams
- Associate messages with the correct request
#### `StreamMessage`
Union type for message-based streaming. Most variants include the `RequestId`:
- `StreamStart(request_id, headers)` - Stream started, initial headers received
- `Chunk(request_id, data)` - Data chunk received (BitArray)
- `StreamEnd(request_id, headers)` - Stream completed, trailing headers received
- `StreamError(request_id, reason)` - Stream failed with error message
- `DecodeError(reason)` - FFI corruption (should be reported as a bug)
**About DecodeError:**
`DecodeError` is a rare error indicating the Erlang→Gleam FFI boundary received
a malformed message from `httpc`. This is **not a normal HTTP error** - it means:
- Erlang/OTP version incompatibility with this library
- Memory corruption or other serious runtime issue
- A bug in this library's FFI code
Unlike other variants, `DecodeError` does not include a `RequestId` because the
request ID itself could not be decoded. If you see this error, please report it
as a bug with the full error message at https://github.com/TrustBound/dream/issues
### Client Configuration (Builder Pattern)
- `client.new` - Create default request configuration
- `client.method(req, method)` - Set HTTP method (GET, POST, etc.)
- `client.scheme(req, scheme)` - Set protocol (HTTP or HTTPS)
- `client.host(req, host)` - Set hostname (required)
- `client.port(req, port)` - Set port number (optional, defaults to 80/443)
- `client.path(req, path)` - Set request path
- `client.query(req, query)` - Set query string
- `client.body(req, body)` - Set request body
- `client.timeout(req, timeout_ms)` - Set request timeout in milliseconds
- `client.add_header(req, name, value)` - Add a single header
- `client.headers(req, headers)` - Replace all headers at once
### Request Execution
#### Blocking Mode
- `client.send(req) -> Result(String, String)`
- Returns complete response body as string
- Blocks until entire response received
- Returns `Ok(body)` on success or `Error(msg)` on failure
#### Yielder Streaming Mode
- `client.stream_yielder(req) -> yielder.Yielder(Result(bytes_tree.BytesTree, String))`
- Returns yielder that produces `Result` values for each chunk
- `Ok(chunk)` - Successful chunk, more may follow
- `Error(reason)` - Terminal error, stream is done
- Normal completion: Yielder returns `Done` (no error yielded)
- Pull-based: blocks while waiting for next chunk
- Use with `yielder.each()`, `yielder.fold()`, `yielder.to_list()`, etc.
- **Not suitable for OTP actors with concurrent operations**
#### Message-Based Streaming Mode
- `client.stream_messages(req) -> Result(RequestId, String)`
- Starts stream and returns `RequestId` immediately
- Push-based: messages sent to your process mailbox automatically
- Non-blocking and fully OTP-compatible
- Returns `Ok(request_id)` if stream started, `Error(msg)` if failed
- `client.select_stream_messages(selector, mapper) -> Selector(msg)`
- Integrates stream messages into an OTP selector
- `mapper` converts `StreamMessage` to your actor's message type
- Use in `actor.Spec` init function to build your selector
- `client.cancel_stream(request_id) -> Nil`
- Cancels an active message-based stream
- No more messages will arrive for this `RequestId`
- Safe to call multiple times on same ID
## Error Handling
All execution modes use `Result` types to force explicit error handling:
### Blocking and Message Start Errors
```gleam
import dream_http_client/client.{send}
import gleam/io.{println_error}
case send(request) {
Ok(body) -> process_response(body)
Error(msg) -> {
// Common errors:
// - Connection refused
// - DNS resolution failed
// - Invalid URL
// - Timeout
println_error("Request failed: " <> msg)
}
}
```
### Streaming Errors
For message-based streaming, errors arrive as `StreamError` messages:
```gleam
import dream_http_client/client.{StreamError}
HttpStream(StreamError(req_id, reason)) -> {
// Handle mid-stream errors:
// - Network interruption
// - Server closed connection
// - Timeout
log_error(req_id, reason)
cleanup_stream(state, req_id)
}
```
### Best Practices
1. **Always handle `Error` cases** - Network operations can fail
2. **Set appropriate timeouts** - Use `client.timeout()` to configure request timeouts (default: 30s)
3. **Handle yielder errors** - `stream_yielder()` produces `Result` values, check each one
4. **Cancel streams when done** - Free resources with `cancel_stream()`
5. **Track active streams** - Use a `Dict(RequestId, State)` in actors
6. **Handle StreamError and DecodeError** - Network can fail mid-stream, FFI can corrupt
7. **Test error paths** - Simulate failures in tests with slow/error endpoints
## Design Principles
This module follows the same quality standards as [Dream](https://github.com/TrustBound/dream):
- **No nested cases** - Clear, flat control flow throughout
- **No anonymous functions** - All functions are named for clarity
- **Builder pattern** - Consistent, composable request configuration
- **Type safety** - `Result` types force error handling at compile time
- **OTP-first design** - Message-based API designed for supervision trees
- **Comprehensive testing** - Unit tests (no network) + integration tests (real HTTP)
- **Battle-tested foundation** - Built on Erlang's production-proven `httpc`
## About Dream
This module was originally built for the [Dream](https://github.com/TrustBound/dream) web toolkit, but it's completely standalone and can be used in any Gleam project. It follows Dream's design principles and will be maintained as part of the Dream ecosystem.
## License
MIT License - see LICENSE file for details.
