The SDK is designed to be integrated into frontend projects, allowing seamless integration with the Keplr wallet extension. It enables users to perform signing and messaging functionalities within the frontend project.
npm install nesa-sdk@https://github.com/nesaorg/nesa-sdk
npm run codegen # proto ==> ts, Generate .ts files into ./src/codecimport { ChatClient } from 'nesa-sdk';
const ChatUtils = new ChatClient({
modelName: "", // model name
chainInfo: {} // optional. chainInfo , Refer to the defaultChainInfo format in src/default.config.ts
walletName: "" // optional. walletName , default keplr , When using metamask, please set this parameter to: npm:@leapwallet/metamask-cosmos-snap
lockAmount: "", // optional. lock amount , default 1000 * tokenPrice
singleLockUpAmount: "" // optional. Number of single signed payment , default 100 * tokenPrice
lowBalance: "" // optional. The remaining unused amount of payment. If unused payment is lower than this value, payment will be regenerated., default 10 * tokenPrice
})
// chainInfo parameter description:
const chainInfo = {
bech32Config: {
bech32PrefixAccAddr: "nesa",
bech32PrefixAccPub: "nesapub",
bech32PrefixConsAddr: "nesavalcons",
bech32PrefixConsPub: "nesavalconspub",
bech32PrefixValAddr: "nesavaloper",
bech32PrefixValPub: "nesavaloperpub",
},
beta: true,
bip44: {
coinType: 118,
},
chainId: "nesa",
chainName: "Nesa Devnet ",
chainSymbolImageUrl:
"https://raw.githubusercontent.com/chainapsis/keplr-chain-registry/main/images/nesa/chain.png",
currencies: [
{
coinDecimals: 6,
coinDenom: "NES",
coinGeckoId: "nesa",
coinImageUrl:
"https://raw.githubusercontent.com/chainapsis/keplr-chain-registry/main/images/nesa/nes.png",
coinMinimalDenom: "unes",
},
],
feeCurrencies: [
{
coinDecimals: 6,
coinDenom: "NES",
coinGeckoId: "nesa",
coinImageUrl:
"https://raw.githubusercontent.com/chainapsis/keplr-chain-registry/main/images/nesa/nes.png",
coinMinimalDenom: "unes",
gasPriceStep: {
average: 0.02,
high: 0.1,
low: 0.01,
},
},
],
nodeProvider: {
email: "[email protected]",
name: "Nesa",
website: "https://nesa.ai/",
},
rest: "https://lcd.dev.nesa.ai",
rpc: "https://rpc.dev.nesa.ai",
stakeCurrency: {
coinDecimals: 6,
coinDenom: "NES",
coinGeckoId: "nesa",
coinImageUrl:
"https://raw.githubusercontent.com/chainapsis/keplr-chain-registry/main/images/nesa/nes.png",
coinMinimalDenom: "unes",
},
}
// If you need to set up a custom access point, you need to refer to the above format and modify the corresponding configuration based on the custom access point. (In particular, please pay attention to the first 5 items of the configuration)
{
chainId: 'nesa', // ChainId for custom access point
chainName: 'Nesa Devnet', // ChainName of custom access point
rest: 'https://lcd.dev.nesa.ai', // Rest address of custom access point
rpc: 'https://rpc.dev.nesa.ai', // Rpc address of custom access point
feeCurrencies: [ // Please configure the feeCurrencies for custom access points to set up transaction fees based on the actual token information of the custom access point.
{
coinDenom: "NES", // Coin Name
coinMinimalDenom: "unes", // Minimum unit of coin
coinDecimals: 6, // Coin precision
coinGeckoId: "nesa", // Coin Name
coinImageUrl:
"https://raw.githubusercontent.com/chainapsis/keplr-chain-registry/main/images/nesa/nes.png", // Coin icon
gasPriceStep: { // Gas price configuration
low: 0.01, // Lowest gas price
average: 0.02, // Average gas price
high: 0.1, // Highest gas price
},
},
],
... // Other chain parameters, can be directly used as reference examples without any special modifications.
}
requestChatStatus is used to return the current conversation progress status, a total of 8 statuses will be returned.
| Code | Message | Remark |
|---|---|---|
| 300 | Connecting to Nesa chain | After the SDK is initialized, it defaults to this state. |
| 301 | Connected to Nesa chain | Nesa chain is connected, and returns after the chain parameters are initialized. (After click 'start inference') |
| 302 | Choosing an inference validator | Signature successful, Choosing an inference validator |
| 303 | Connecting to the validator | Connecting to the validator |
| 304 | Waiting for query | Waiting for query |
| 305 | Conducting inference | Conducting inference (After click 'start query') |
| 306 | Receiving responses | Receiving responses |
| 307 | Task completed, wait for another query | Task completed, wait for another query |
This is a promise that will call back a readableStream object. You can get the conversation information through readableStream.on('data'), which will return an object:
{
code: 200, // code
message: "message", // message
}
Return Code and message
| Code | Message | Remark |
|---|---|---|
| 200 | TransactionHash | requestSession transaction Hash |
| 311 | LockAmount cannot be less than x | LockAmount check |
| 312 | Sign register session failed | Sign register session failed message |
| 313 | Register session error | Register session failed error |
| 314 | Chain params format error | Chain params format error |
| 315 | Chain params error | Chain params error message |
| 316 | SDK client init error | SDK client init error |
| 317 | Wallet connect error | Wallet connect error message |
| 318 | Broadcast tx error | Broadcast tx error message |
| 319 | Agent connection error | Agent connection error message |
This interface is used for initiating a conversation, with the parameters:
{
"messages": [
// ...you can add history messages
// Please follow the following rules to set role:
// Use "user" for user queries (unless you know that the model side uses a different name);
// The content of the assistant's answer is always returned using "assistant".
// For example:
{'role': 'user', 'content': 'user question 1'},
{'role': 'assistant', 'content': 'assistant answer 1'},
{'role': 'user', 'content': 'user question 2'},
{'role': 'assistant', 'content': 'assistant answer 2'},
...
],
"model": "", // Model name
... // Other parameters, you need to ensure that these parameters are supported by the LLM Backend
}
This is a promise that will call back a readableStream object. You can get the conversation information through readableStream.on('data'), which will return an object:
{
code: 200, // code
message: "hello", // message
total_payment: { // total_payment , If code is 200 it will return
amount: 10, // total_payment amount
denom: 'unes' // denom
},
session_id: "" // session_id
}
Return Code and message
| Code | Message | Remark |
|---|---|---|
| 200 | Normal response | |
| 201 | No signature found or the signature has expired. | |
| 202 | Illegal link | |
| 203 | Current chat contributions | |
| 204 | websocket connection error message |
|
| 205 | Business error information returned |
This method is used to actively close the heartbeat connection between the SDK and the agent, thereby ending the conversation.
Explanation of the heartbeat connection between the SDK and agent:
-
After the SDK is initialized successfully and the session is successfully registered using the requestSession method, the SDK will actively maintain a heartbeat connection with the agent. As long as the heartbeat connection is normal (of course, the token consumption of the conversation cannot exceed the locked amount), the agent will continue to provide normal conversation functionality. Once the heartbeat connection is closed (either actively by calling the requestCloseHeartbeat method, or due to network issues causing the connection to be closed) for more than 20 seconds, the agent service will terminate the current conversation.
-
If no conversation is initiated after the SDK is initialized successfully and the session is successfully registered using the requestSession method, the token locked when requesting the session will be returned to the original account by the chain after the timeout period configured in the chain (currently set to 4 hours).
-
If a conversation is initiated after the SDK is initialized successfully and the session is successfully registered using the requestSession method, but not all of the locked tokens are consumed, the chain will return the unused tokens to the original account when the agent service terminates the conversation.
- Setting up the Keplr wallet plugin:
Path: Settings -> Advanced -> Modify connection point
Choose: Nesa Devnet, set
RPC: https://rpc.dev.nesa.ai
LCD: https://lcd.dev.nesa.ai
-
Parameter settings when initiating a conversation
The SDK accepts any parameter values. Please pass parameters in key-value format when initiating a conversation using the "requestChat" method. There are no specific data type requirements for the values, and there is no limit to the number of parameters. However, please ensure that these parameters are supported by the LLM backend. You can refer to the following implementation for guidance.
import { ChatClient } from 'nesa-sdk';
const ChatUtils = new ChatClient({
modelName: "",
chainInfo: ChainInfo // optional, The default chain config is src/default.config.ts,
lockAmount: lockAmount // optional, default is 1000
})
ChatUtils.requestChat({
"messages": [
{'role': 'user', 'content': 'user question 1'},
{'role': 'assistant', 'content': 'assistant answer 1'},
...
],
"model": "", // model name
// you can add other parameter,like:
"frequency_penalty": "", // Optional frequency_penalty
"presence_penalty": "" // Optional presence_penalty
"session_id": "", // optional, session_id
})
.then(readableStream => {
readableStream.on("data",(data) => {
// Processing transmission data
const {code, message, total_payment, session_id} = data
})
readableStream.on("end",() => {
// End of transmission
})
})
- Test Cases
// nodejs version >= v19.0.0
node test/test.js
import { ChatClient } from 'nesa-sdk';
const ChatUtils = new ChatClient({
modelName: "",
chainInfo: ChainInfo // optional, The default chain config is src/default.config.ts,
lockAmount: lockAmount // optional, default is 1000000
})
// This method can be called once
ChatUtils.requestSession()
.then(result => {
readableStream.on("data",(data) => {
// Processing transmission data
const {code, message} = data
if (code === 200) {
// Streaming data return for TransactionHash
const transactionHash = message
...
} else {
// Exception information prompt
const progressMessage = message
...
}
// For detailed code and message, please refer to the above API
// 200 : requestSession transaction Hash
// 311 : LockAmount cannot be less than x, LockAmount check
// 312 : Sign register session failed
// 313 : Register session error
// 314 : Chain params format error
// 315 : Chain params error
// 316 : SDK client init error
// 317 : Wallet connect error
})
readableStream.on("end",() => {
// End of transmission
})}
})
.catch(error => {
// Error handling
});
// Please call ChatUtils.requestSession before calling and successfully get the callback
ChatUtils.requestChat({
"messages": [
// ...you can add history messages
// Please follow the following rules to set role:
// Use "user" for user queries (unless you know that the model side uses a different name);
// The content of the assistant's answer is always returned using "assistant".
// For example:
{'role': 'user', 'content': 'user question 1'},
{'role': 'assistant', 'content': 'assistant answer 1'},
{'role': 'user', 'content': 'user question 2'},
{'role': 'assistant', 'content': 'assistant answer 2'},
...
],
"model": "", // model name
// you can add other parameter,like:
"frequency_penalty": 0.5, // optional frequency_penalty
"session_id": "", // optional, session_id
})
.then(readableStream => {
readableStream.on("data",(data) => {
// Processing transmission data
const {code, message, total_payment, session_id} = data
if (code === 200) {
// Streaming data return for normal conversation
// session_id will only be returned when the code is 200.
const nextChatResponse = message
const totalPayment = total_payment
...
} else {
// Exception information prompt
const infoMessage = message
...
}
// For detailed code and message, please refer to the above API
// 200 : Normal data return
// 201 : No signature found or the signature has expired
// 202 : Illegal agent websocket link
// 203 : Current query contributions
// 204 : Websocket connection abnormality error message
// 205 : Websocket business error information returned
})
readableStream.on("end",() => {
// End of transmission
})
})
.catch(error => {
// Error handling
});
// This interface is used to retrieve the status of the conversation progress.
ChatUtils.requestChatStatus()
.then(readableStream => {
readableStream.on("data",(data) => {
// Processing transmission data
const {code, message} = data
// code and message reference as follows
// 300 : Connecting to Nesa chain
// 301 : Connected to Nesa chain
// 302 : Choosing an inference validator
// 303 : Connecting to the validator
// 304 : Waiting for query
// 305 : Conducting inference
// 306 : Receiving responses
// 307 : Task completed, wait for another query
})
readableStream.on("end",() => {
// End of transmission
})
})
.catch(error => {
// Error handling
});
// This method is used to disconnect the heartbeat connection, thereby disconnecting the conversation and triggering the agent to submit payment.
// Please be aware that this method should only be called when it is necessary to end the conversation.
ChatUtils.requestCloseHeartbeat()