The Tekimax SDK provides a robust middleware architecture via Plugins.
Instead of wrapping the entire SDK or duplicating logic across your app to handle caching, redaction, or logging, you can plug these directly into the core Tekimax client.
Pre-built Plugins
The SDK exports three core plugins out of the box used for Enterprise readiness:
LoggerPlugin(Telemetry): Automatically logs requests, stream chunks, tokens, and tool execution boundaries.PIIFilterPlugin(Security): Redacts standard patterns like emails ([REDACTED EMAIL]) and SSNs before they hit the provider.MaxContextOverflowPlugin(Scalability): Prevents token bloat in long-running loops by gently truncating the oldest messages (while safely preserving your System prompt).
Usage
Pass the plugins array when initializing your Tekimax client:
Code
import {
Tekimax,
OpenAIProvider,
PIIFilterPlugin,
LoggerPlugin,
MaxContextOverflowPlugin
} from 'tekimax-ts'
const client = new Tekimax({
provider: new OpenAIProvider({ apiKey: 'sk-...' }),
plugins: [
new LoggerPlugin(),
new PIIFilterPlugin(),
new MaxContextOverflowPlugin(15) // Keeps a rolling window of 15 messages max
]
})Building Custom Plugins
Building a custom plugin is as simple as implementing the TekimaxPlugin interface.
You can hook into the exact moment before a request is sent, after a response arrives, during every stream chunk, or around tool executions.
The TekimaxPlugin Interface
Code
import type { PluginContext, ChatResult, StreamChunk, TekimaxPlugin } from 'tekimax-ts'
export interface TekimaxPlugin {
name: string
/** Triggered when the Tekimax client is instantiated */
onInit?: (client: any) => void
/** Triggered before a chat or stream request is sent. Can mutate the context. */
beforeRequest?: (context: PluginContext) => Promise<void | PluginContext>
/** Triggered after a fully completed standard chat response */
afterResponse?: (context: PluginContext, result: ChatResult) => Promise<void>
/** Triggered on every chunk during a streaming response */
onStreamChunk?: (context: PluginContext, chunk: StreamChunk) => void
/** Triggered before a tool is executed */
beforeToolExecute?: (toolName: string, args: unknown) => Promise<void>
/** Triggered after a tool is executed */
afterToolExecute?: (toolName: string, result: unknown) => Promise<void>
}Example: Custom Langfuse Logger
Here is an example of how you might integrate a third-party observability platform like Langfuse using the plugin architecture:
Code
export class LangfusePlugin implements TekimaxPlugin {
name = 'LangfusePlugin'
private langfuse = new Langfuse(...)
async beforeRequest(context: PluginContext) {
// Start a Langfuse observation trace
this.langfuse.trace({
name: "Chat Generation",
model: context.model,
input: context.messages
})
}
async afterResponse(context: PluginContext, result: ChatResult) {
// Log the completion and token usage
this.langfuse.generation({
output: result.message.content,
usage: result.usage
})
}
}