📖 Documentation

Read the guides over at Chainflow Docs to get started!

⚠️ (Important!) Note

Chainflow is my first and somewhat experimental open-source project which I started to assist my own dev work. It's similar to Postman Flows, but done in code so us engineers have more control to weave logic in between API endpoint calls (+ the ease of writing simple scripts with Chainflow to be run in your CLI or as pipeline jobs etc.).

I am still finding and ironing out pain points in using the library and releases should be considered unstable while I improve the library's usability. It is far from perfect in its current state, but I will continue to improve upon it if I or others find potential in its use. Therefore, your feedback and ideas are very important to me. I welcome thoughts on just about anything - useful dev tools to manage this project, QoL features to support easier coding patterns when using this library or even thoughts on the future direction of this project.

Drop them as a Github issue or email me at [email protected]!

💭 When might Chainflow be useful?

1. Setting up demo data

   Say you have an app that you're developing new features for and you'd like to demo those features. You need your app to be in a certain context and your database in a specific state - perhaps that context requires that a user has logged in with certain permissions, has created a "group" in the app and has added other users to that group. To create that context, you could manually click through your app to log in as a user and set everything up, but that could be tedious. Alternatively, you may use raw SQL or other DB scripts to directly manipulate the DB and insert users, roles, etc. However, those scripts could miss out on important side effects relevant to the business context of your app - side effects that tend to be triggered by services exposed by your backend server.

   Instead, you can setup the context in your app by calling the relevant service endpoints you have built (POST /user, POST /role, etc.) and triggering their business logic to make your data setup adhere closely to the way your app truly behaves - as if a user was really logging in and doing everything. Chainflow helps you do this by providing tools to compose API call workflows. You can then avoid manual setups and minimize your use of database scripts to just data that is not configurable with existing endpoints!

2. Speeding up development

   Similar to setting up demo data, while coding new features you may often want to test out how they behave in your app, and again you may want your app to be in a specific state locally for that. You can write API call workflow scripts built with Chainflow to help move your app into those states quickly.

3. Testing your endpoints

   An API call workflow could behave as if it were a frontend client calling the backend. In that way, you can run UI-agnostic end-to-end testing of backend endpoints by using API call workflows to simulate how a frontend would interact with the backend.

Basic Usage

Note: Chainflow may not be compatible with Node versions past end-of-life (v16.x and below).

npm install --save-dev chainflow

origin creates a template on which you define your API endpoints by calling the endpoint's HTTP method verb (e.g. post, patch) as methods.

import { origin } from chainflow;

const backend = origin('127.0.0.1:5000');

const createUser = backend.post('/user').body({
  name: 'Tom',
  details: {
    age: 40,
  },
});

const createRole = backend.post('/role').body({
  type: 'Engineer',
  userId: createUser.resp.body.id,
});

const getUser = backend.get('/user').query({
  roleType: createRole.resp.body.type,
});

Use a chainflow to define a sequence of endpoint calls that take advantage of the values and links provided above.

import { chainflow } from Chainflow;

const flow = chainflow()
  .call(createUser)
  .call(createRole)
  .call(getUser);

flow.run();

The above setup will result in the following endpoint calls:

1. POST Request to /user with body:

   {
     "name": "Tom",
     "details": {
       "age": 40
     }
   }

2. POST Request to /role with body:

   {
     "type": "Engineer",
     "userId": "['userId' from response to step 1]"
   }

3. GET Request to /user?roleType=['type' from response to step 2]

More Features

Query params

Define query params with the query method on an endpoint.

const getUsersInGroup = backend.get('/user').query({ groupId: 'some-id' });

Path params

Define path params by wrapping a param name with {} in the endpoint path.

const getGroupsWithUser = backend.get('/groups/{userId}');

You can specify values for your path params by calling pathParams. Note that path params which do not actually exist in the path will be discarded.

const getGroupsWithUser = backend.get('/groups/{userId}').pathParams({
  userId: 'user123',
});

Headers

Specify headers with headers method on endpoints.

const getInfo = backend.get('/info').headers({ token: 'some-token' });

You can also use headers on an Origin to have all endpoints made for that origin bear those headers.

const backend = origin('127.0.0.1:3001').headers({ token: 'some-token' });

const getInfo = backend.get('/info'); // getInfo endpoint will have the headers defined above

Default headers

Chainflow attaches default headers to all requests made by any endpoint with the value:

'content-type': 'application/json',
'User-Agent': 'Chainflow/[major.minor version number]',

If you'd like to change this, pass your default headers to the defaultHeaders util.

import { defaultHeaders } from 'chainflow';

defaultHeaders({ 'content-type': 'application/xml' });

Pass in true as the second argument if you want to replace the entire set of default headers. Otherwise, the example above only overwrites the content-type default header and keeps User-Agent.

Initializing Values

The request payloads under Basic Usage are defined with only default values - i.e. the values which a Chainflow use if there are no response values from other endpoint calls linked to it.

However, you can also use the following features to more flexibly define the values used in a request.

required

Marks a value as required but without a default. The chainflow will expect this value to be sourced from another node. If no such source is available, the endpoint call will throw an error.

const createUser = backend.post('/user').body({
  name: required(),
});

gen

Provide a callback that generates values for building requests.

const randAge = () => Math.floor(Math.random() * 100);

const createUser = backend.post('/user').body({
  name: 'Tom',
  details: {
    age: gen(randAge),
  },
});

link

You can use the link function to specify a callback to transform the response value before it is passed to the input node.

const addGreeting = (name: string) => `Hello ${name}`;

const createMessage = backend.post('message').body({
  msg: link(getUser.resp.body.name, addGreeting);
});

set

The link has another function signature.

You can use the set method on an endpoint to expose its input nodes, then use the 2nd function signature of link as shown below: pass in the input node first (msg), then the source node second and optionally a callback third.

createMessage.set(({ body: { msg } }) => {
  link(msg, getUser.resp.body.name);
  link(msg, createUser.resp.body.name);
});

With a callback:

createMessage.set(({ body: { msg } }) => {
  link(msg, getUser.resp.body.name, addGreeting);
  link(msg, createUser.resp.body.name, addGreeting);
});

linkMerge

Link multiple response values to a single request node with an optional callback to merge the values into a single input value. This has 4 function signatures:

For the argument containing the source nodes, you can either pass an array of SourceNodes:

// note the callback has an array parameter
const mergeValues = ([name, favAnimal]: [string, string]) =>
  `${name} likes ${favAnimal}.`;

const createMessage = backend.post('message').body({
  msg: linkMerge(
    // array of source nodes
    [getUser.resp.body.name, getFavAnimal.resp.body.favAnimal],
    mergeValues,
  );
});

or you can pass an object with SourceNodes as the values:

// note the callback has an object parameter
const mergeValues = ({
  userName,
  favAnimal,
}: {
  userName: string;
  favAnimal: string;
}) => `${userName} likes ${favAnimal}.`;


const createMessage = backend.post('message').body({
  msg: linkMerge(
    // object of source nodes
    {
      userName: getUser.resp.body.name,
      favAnimal: getFavAnimal.resp.body.favAnimal,
    },
    mergeValues,
  );
});

alternatively, you can use the set method in addition with the other function signature of linkMerge (similar to how link above has overloads to work with set).

with array:

createMessage.set(({ body: { msg } }) => {
  linkMerge(
    msg, // the input node
    [getUser.resp.body.name, getFavAnimal.resp.body.favAnimal],
    mergeValues,
  );
});

with object:

createMessage.set(({ body: { msg } }) => {
  linkMerge(
    msg, // the input node
    {
      userName: getUser.resp.body.name,
      favAnimal: getFavAnimal.resp.body.favAnimal,
    },
    mergeValues,
  );
});

Note that the merging link created by this method will only be used if ALL the source nodes specified are available i.e. if either one of getUser.resp.body.name or getFavAnimal.resp.body.favAnimal does not have a value, this link will not be used at all.

Call Options

You can declare manual values for an endpoint call in the chainflow itself, should you need to do so, by passing in a Call Options object as a second argument in the call method.

body, pathParams, query and headers can be set this way.

const createUser = backend.post('/user').body({
  name: 'Tom',
});

chainflow()
  .call(createUser, { body: { name: 'Harry' } })
  .run();

seed

You can specify request nodes to take values from the chainflow 'seed' by importing the seed object and linking nodes to it. Provide actual seed values by calling the seed method on a chainflow before you run it, like below.

import { chainflow, link seed, } from 'chainflow';

const createUser = backend.post('/user').body({
  name: required(),
});

createUser.set(({ body: { name }}) => {
  link(name, seed.username);
});

chainflow()
  .call()
  .seed({ username: 'Tom' })
  .run();

Allow Undefined Sources Values

By default, an input node will reject and skip a source node's value if it is unavailable or undefined. However, you can change this by passing a source node into the config utility function and passing an options object as the second parameter like below. This informs an input node to use the source node's value regardless of whether the value is undefined or not.

import { config } from 'chainflow';

createUser.set(({ body: { name } }) => {
  link(name, config(seed.username, { allowUndefined: true }));
});

This has important implications - it means that as long as the source (e.g. a response from an endpoint call) is available, then the linked source node's value will be taken and used (even if that value is unavailable, which would be taken as undefined). Therefore, any other linked sources will not be used UNLESS 1. they have a higher priority or 2. the source providing the linked node that allows undefined is unavailable.

clone

You can create chainflow "templates" with the use of clone to create a copy of a chainflow and its endpoint callqueue. The clone can have endpoint calls added to it without modifying the initial flow.

const initialFlow = chainflow().call(endpoint1).call(endpoint2);

const clonedFlow = initialFlow.clone();

clonedFlow.call(endpoint3).run(); // calls endpoint 1, 2 and 3
initialFlow.call(endpoint4).run(); // calls endpoint 1, 2 and 4

extend

You can connect multiple different chainflows together into a longer chainflow using extend.

const flow1 = chainflow().call(endpoint1).call(endpoint2);
const flow2 = chainflow().call(endpoint3);

flow1.extend(flow2).run(); // calls endpoint 1, 2 and 3

config

respParser
By default, a chainflow parses response bodies as JSON objects UNLESS the status code is 204 or the content-type header does not contain application/json (to avoid errors when parsing an empty body), upon which they will instead parse it as text.

To set a specific parsing format, you can call .config to change that configuration on an endpoint (or on an Origin, to apply it to all endpoints created from it) like so:

import { RESP_PARSER } from 'chainflow';

const getUser = backend.get('/user').config({
  respParser: RESP_PARSER.TEXT,
});

or with camelcase in JavaScript:

const getUser = backend.get('/user').config({
  respParser: 'text',
});

There are 4 supported ways to parse response bodies (provided by the underlying HTTP client undici): arrayBuffer, blob, json and text.

respValidator
Another configuration option is how to validate the response to an endpoint. By default, Chainflow rejects responses that have HTTP status code 400 and above and throws an error. You can pass in a custom respValidator to change when a response is rejected.

const getUser = backend.get('/user').config({
  respValidator: (resp) => {
    if (resp.statusCode !== 201) return { valid: false, msg: 'Failed to retrieve users.' };
    if (!Object.keys(resp.body as Record<string, unknown>).includes('id'))
      return { valid: false, msg: 'Response did not provide user ID.' };
    return { valid: true };
  },
});

Your custom validator callback should have a return type:

{
  valid: boolean; // false if response should be rejected
  msg?: string; // error message
}

store

Instead of direct links between endpoints, you can use a central store to keep values from some endpoints and have other endpoints take from it via the special store object.

import { store } from 'chainflow';

const createUser = backend
  .post('/user')
  .body({
    name: 'Tom',
  })
  .store((resp) => ({
    // this endpoint will store `id` from a response to `userId` in the store
    userId: resp.body.id,
  }));

const addRole = backend.post('/role').body({
  // this endpoint will take `userId` from the store, if available
  userId: store.userId,
  role: 'Engineer',
});

chainflow().call(createUser).call(addRole).run();

This is usually useful when you have endpoints that could take a value from any one of many other endpoints for the same input node. Having a store to centralise these many-to-many relationships (like an API Gateway) can improve the developer experience.

continuesFrom - transferring Chainflow states

Say we have 2 endpoints, login and createGroup. We want to login as a user once, then proceed to proceed 3 groups as that same user without having to login 3 times.

const createGroup = backend
  .post('/group')
  .headers({
    Authorization: login.resp.body.authToken,
  })
  .body({
    groupName: seed.groupName,
  });

// loggedInFlow will contain a response from the `login` endpoint
const loggedInFlow = chainflow().call(login).run();

// createGroupFlow will take the response that
// loggedInFlow received and carry on from there
const createGroupFlow = chainflow().call(createGroup).continuesFrom(loggedInFlow);

const groupNames = ['RapGPT', 'Averageexpedition', 'Shaky Osmosis'];
for (const groupName in groupNames) {
  createGroupFlow.seed({ groupName }).run();
}

We run a chainflow that calls login first to get a response from the login endpoint.

Using the continuesFrom method, createGroupFlow will copy the state of source values (i.e. responses) from loggedInFlow. This means createGroupFlow will now have the logged in user's authToken received from calling login, and will use it when calling createGroup thrice for each group name in the groupNames array.

events

After running a chainflow, you can retrieve the request and response event data from endpoint calls executed during that run via the events property on that chainflow.

const flow = chainflow().run(createUser).run(getRoles);

const events = flow.events;

The events will look something like:

[
  {
    details: '[POST] /user', // identifies the endpoint called
    req: {
      method: 'POST',
      url: ...,
      body: ...,
      headers: ...,
      respParser: ..., // the format used to parse the response body
    },
    resp: {
      statusCode: 200,
      body: ...,
      headers: ...,
      ...
    }
  },
  {
    details: '[GET] /roles',
    req: ...,
    resp: ...
  }
]

The responses in the array follow the order in which the respective endpoints are called.

logging

Log requests/responses in your console by setting ENABLE_CHAINFLOW_LOGS=true in your environment variables, or by simply importing and calling the enableLogs function.

import { enableLogs } from 'chainflow';

enableLogs();

Misc Behaviors

• If you have multiple endpoint calls to the same endpoint on one chainflow and they are linked to other endpoints' input nodes further down the flow, the latest endpoint call's values will be used.

For example:

chainflow()
  .call(getUser) // 1st call
  .call(addRole)
  .call(getUser) // 2nd call
  .call(createGroup);

If an input node on createGroup requires a value from a response to getUser, then createGroup will take that value from the latest call to getUser (i.e. from the response to the 2nd call to getUser after the call to addRole).

Future Updates

Below features are currently not yet supported but are planned in future releases.

1. More flexibility to log and return responses
2. Conditional calls - execute an endpoint call only if some condition is met
3. (Exploratory) API performance measurement
4. (Exploratory) Possibly some sort of UI/diagram generation

Development

Areas that could be better (non-exhaustive)

Encoding endpoint IDs

• Currently assumes that URLs of endpoints do not contain unencoded | and [] characters. [] used to wrap around HTTP method in the encoded ID. Linkmerge uses | to separate different encoded IDs.
• Current implementation also leads to ID collision if multiple endpoints with the same method and path are created (but perhaps with different configuration) and are called on the same chainflow.
• Idea: Have a centralized service to issue unique IDs to deconflict endpoints - but still somehow encode the method/path info of an endpoint into it.

Logging

• Should further explore appropriate degree of detail for logging
• Truncation of requests/responses with extremely large payloads

Proxy performance

• Creating proxies recursively when defining SourceNode paths could hit performance - not rigorously tested yet, so not 100% sure of how significant it is

Trivia

• You probably noticed that I enjoy using the Builder pattern for its clarity.
• I'm praying the wave 🌊 emoji remains sufficiently shaped like a "C" to avoid confusion. Please let me know if there is some system where it does not!