222 lines
8.2 KiB
JavaScript
222 lines
8.2 KiB
JavaScript
import { BaseCallbackHandler } from "@langchain/core/callbacks/base";
|
|
/**
|
|
* Upstash Ratelimit Error
|
|
*
|
|
* Raised when the rate limit is reached in `UpstashRatelimitHandler`.
|
|
*/
|
|
class UpstashRatelimitError extends Error {
|
|
/**
|
|
* @param message - Error message
|
|
* @param type - The kind of limit which was reached. One of "token" or "request"
|
|
* @param limit - The limit which was reached. Passed when type is request
|
|
* @param reset - Unix timestamp in milliseconds when the limits are reset. Passed when type is request
|
|
*/
|
|
constructor(message, type, limit, reset) {
|
|
super(message);
|
|
Object.defineProperty(this, "type", {
|
|
enumerable: true,
|
|
configurable: true,
|
|
writable: true,
|
|
value: void 0
|
|
});
|
|
Object.defineProperty(this, "limit", {
|
|
enumerable: true,
|
|
configurable: true,
|
|
writable: true,
|
|
value: void 0
|
|
});
|
|
Object.defineProperty(this, "reset", {
|
|
enumerable: true,
|
|
configurable: true,
|
|
writable: true,
|
|
value: void 0
|
|
});
|
|
this.type = type;
|
|
this.limit = limit;
|
|
this.reset = reset;
|
|
}
|
|
}
|
|
/**
|
|
* Callback to handle rate limiting based on the number of requests
|
|
* or the number of tokens in the input.
|
|
*
|
|
* It uses Upstash Ratelimit to track the rate limit which utilizes
|
|
* Upstash Redis to track the state.
|
|
*
|
|
* Should not be passed to the chain when initializing the chain.
|
|
* This is because the handler has a state which should be fresh
|
|
* every time invoke is called. Instead, initialize and pass a handler
|
|
* every time you invoke.
|
|
*/
|
|
class UpstashRatelimitHandler extends BaseCallbackHandler {
|
|
/**
|
|
* @param identifier - The identifier to rate limit, like a user ID or an IP address
|
|
* @param options - Ratelimit options
|
|
*/
|
|
constructor(identifier, options) {
|
|
super();
|
|
Object.defineProperty(this, "name", {
|
|
enumerable: true,
|
|
configurable: true,
|
|
writable: true,
|
|
value: "UpstashRatelimit"
|
|
});
|
|
Object.defineProperty(this, "raiseError", {
|
|
enumerable: true,
|
|
configurable: true,
|
|
writable: true,
|
|
value: true
|
|
});
|
|
Object.defineProperty(this, "_checked", {
|
|
enumerable: true,
|
|
configurable: true,
|
|
writable: true,
|
|
value: false
|
|
});
|
|
Object.defineProperty(this, "identifier", {
|
|
enumerable: true,
|
|
configurable: true,
|
|
writable: true,
|
|
value: void 0
|
|
});
|
|
Object.defineProperty(this, "tokenRatelimit", {
|
|
enumerable: true,
|
|
configurable: true,
|
|
writable: true,
|
|
value: void 0
|
|
});
|
|
Object.defineProperty(this, "requestRatelimit", {
|
|
enumerable: true,
|
|
configurable: true,
|
|
writable: true,
|
|
value: void 0
|
|
});
|
|
Object.defineProperty(this, "includeOutputTokens", {
|
|
enumerable: true,
|
|
configurable: true,
|
|
writable: true,
|
|
value: void 0
|
|
});
|
|
Object.defineProperty(this, "llmOutputTokenUsageField", {
|
|
enumerable: true,
|
|
configurable: true,
|
|
writable: true,
|
|
value: void 0
|
|
});
|
|
Object.defineProperty(this, "llmOutputTotalTokenField", {
|
|
enumerable: true,
|
|
configurable: true,
|
|
writable: true,
|
|
value: void 0
|
|
});
|
|
Object.defineProperty(this, "llmOutputPromptTokenField", {
|
|
enumerable: true,
|
|
configurable: true,
|
|
writable: true,
|
|
value: void 0
|
|
});
|
|
if (!options.tokenRatelimit && !options.requestRatelimit) {
|
|
throw new Error("You must pass at least one of tokenRatelimit or requestRatelimit.");
|
|
}
|
|
this.identifier = identifier;
|
|
this.tokenRatelimit = options.tokenRatelimit;
|
|
this.requestRatelimit = options.requestRatelimit;
|
|
this.includeOutputTokens = options.includeOutputTokens ?? false;
|
|
this.llmOutputTokenUsageField =
|
|
options.llmOutputTokenUsageField ?? "tokenUsage";
|
|
this.llmOutputTotalTokenField =
|
|
options.llmOutputTotalTokenField ?? "totalTokens";
|
|
this.llmOutputPromptTokenField =
|
|
options.llmOutputPromptTokenField ?? "promptTokens";
|
|
}
|
|
/**
|
|
* Run when the chain starts running.
|
|
*
|
|
* This method is called multiple times during a chain execution.
|
|
* To ensure it only runs once, it checks and updates a `_checked` state.
|
|
*
|
|
* @param _chain - Serialized chain
|
|
* @param _inputs - Chain input values
|
|
* @throws UpstashRatelimitError - If the request rate limit is reached
|
|
*/
|
|
async handleChainStart(_chain, _inputs) {
|
|
if (this.requestRatelimit && !this._checked) {
|
|
const response = await this.requestRatelimit.limit(this.identifier);
|
|
if (!response.success) {
|
|
throw new UpstashRatelimitError("Request limit reached!", "request", response.limit, response.reset);
|
|
}
|
|
this._checked = true;
|
|
}
|
|
}
|
|
/**
|
|
* Run when the LLM starts running.
|
|
*
|
|
* @param _llm - Serialized LLM
|
|
* @param _prompts - Prompts passed to the LLM
|
|
* @throws UpstashRatelimitError - If the token rate limit is reached
|
|
*/
|
|
async handleLLMStart(_llm, _prompts, _runId, _parentRunId, _extraParams, _tags, _metadata, _name) {
|
|
if (this.tokenRatelimit) {
|
|
const result = await this.tokenRatelimit.getRemaining(this.identifier);
|
|
// result of getRemaining was changed from a number to an object in v2.0.0.
|
|
// we check to make sure that it works with versions before & after:
|
|
const remaining = typeof result === "number" ? result : result.remaining;
|
|
if (remaining <= 0) {
|
|
throw new UpstashRatelimitError("Token limit reached!", "token");
|
|
}
|
|
}
|
|
}
|
|
/**
|
|
* Run when the LLM ends running.
|
|
*
|
|
* If the `includeOutputTokens` is set to true, the number of tokens
|
|
* in the LLM completion are counted for rate limiting.
|
|
*
|
|
* @param output - LLM result output
|
|
* @throws Error - If the LLM response does not include required token usage information
|
|
*/
|
|
async handleLLMEnd(output, _runId, _parentRunId, _tags) {
|
|
if (this.tokenRatelimit) {
|
|
const llmOutput = output.llmOutput || {};
|
|
try {
|
|
const tokenUsage = llmOutput[this.llmOutputTokenUsageField];
|
|
const tokenCount = this.includeOutputTokens
|
|
? tokenUsage[this.llmOutputTotalTokenField]
|
|
: tokenUsage[this.llmOutputPromptTokenField];
|
|
if (tokenCount !== undefined) {
|
|
await this.tokenRatelimit.limit(this.identifier, {
|
|
rate: tokenCount,
|
|
});
|
|
}
|
|
else {
|
|
throw new Error("tokenCount not found in llm output");
|
|
}
|
|
}
|
|
catch (error) {
|
|
// eslint-disable-next-line no-instanceof/no-instanceof
|
|
if (error instanceof UpstashRatelimitError) {
|
|
throw error;
|
|
}
|
|
console.error(`Failed to log token usage for Upstash rate limit. It could be because the LLM returns the token usage in a different format than expected. See UpstashRatelimitHandler parameters. Got error: ${error}`);
|
|
}
|
|
}
|
|
}
|
|
/**
|
|
* Creates a new UpstashRatelimitHandler object with the same
|
|
* ratelimit configurations but with a new identifier if it's
|
|
* provided.
|
|
*
|
|
* Also resets the state of the handler.
|
|
*
|
|
* @param identifier - Optional new identifier to use for the new handler instance
|
|
* @returns New UpstashRatelimitHandler instance
|
|
*/
|
|
reset(identifier) {
|
|
return new UpstashRatelimitHandler(identifier ?? this.identifier, {
|
|
tokenRatelimit: this.tokenRatelimit,
|
|
requestRatelimit: this.requestRatelimit,
|
|
includeOutputTokens: this.includeOutputTokens,
|
|
});
|
|
}
|
|
}
|
|
export { UpstashRatelimitHandler, UpstashRatelimitError, };
|