Tekimax organizes AI capabilities into modalities. Each modality maps to a namespace on the client. Providers opt into capabilities explicitly — calling an unsupported modality throws a compile-time TypeScript error, not a runtime 404.
Text
Chat, streaming, embeddings, tool calling
Images
Generation (DALL-E) and vision analysis
Audio
Text-to-speech and transcription (Whisper)
Video
Analysis via Gemini (generation coming soon)
Text
The client.text namespace handles all language model interactions.
import { Tekimax, OpenAIProvider } from 'tekimax-omat';
const client = new Tekimax({ provider: new OpenAIProvider({ apiKey: process.env.OPENAI_API_KEY! }) });
// Standard chat
const response = await client.text.chat.completions.create({
model: 'gpt-4o',
messages: [{ role: 'user', content: 'Summarize this grant report' }],
});
console.log(response.message.content);
// Streaming
const stream = client.text.chat.completions.createStream({
model: 'gpt-4o',
messages: [{ role: 'user', content: 'Write a program description' }],
});
for await (const chunk of stream) {
process.stdout.write(chunk.delta);
}
// Embeddings
const embedding = await client.text.embed({
model: 'text-embedding-3-small',
input: 'Workforce development for returning citizens',
});
console.log(embedding.data[0].embedding); // number[]Images
Image Generation
Supported by: OpenAI (DALL-E 3), OpenRouter (model-dependent)
const result = await client.images.generate({
prompt: 'A diverse group of young adults collaborating on a coding project, warm lighting',
model: 'dall-e-3',
size: '1024x1024',
});
console.log(result.data[0].url);Image Analysis (Vision)
Analyze images with multi-modal models. The SDK normalizes the format — OpenAI uses image_url content parts, Anthropic uses image source blocks, Gemini uses inlineData — but you always call the same method.
Supported by: OpenAI (GPT-4o), Anthropic (Claude 3.x), Gemini, OpenRouter
const analysis = await client.images.analyze({
model: 'gpt-4o',
image: 'https://example.org/student-drawing.png', // URL or base64 data URL
prompt: 'Describe what this student has drawn. Note any labels or text visible.',
});
console.log(analysis.content);OMAT use case: The AssessmentPipeline uses vision analysis to process student drawings and handwriting submissions — the image is described into text before rubric scoring. See the Assessment Guide.
Audio
Text-to-Speech
Supported by: OpenAI
const audio = await client.audio.speak({
model: 'tts-1',
input: 'Welcome to the program. Let\'s get started with your intake assessment.',
voice: 'nova', // alloy | echo | fable | onyx | nova | shimmer
});
// Returns ArrayBuffer — write to file or pipe to audio player
await fs.promises.writeFile('welcome.mp3', Buffer.from(audio.buffer));Transcription (Speech-to-Text)
Supported by: OpenAI (Whisper)
import fs from 'node:fs';
const transcription = await client.audio.transcribe({
file: fs.readFileSync('participant-response.mp3'),
model: 'whisper-1',
language: 'es', // Optional: ISO 639-1 language code
response_format: 'verbose_json', // Includes timestamps and segments
});
console.log(transcription.text);
console.log(transcription.segments); // [{ start, end, text }, ...]OMAT use case: The AssessmentPipeline transcribes speech responses automatically before rubric scoring. Set modality: 'speech' on your StudentResponse and the pipeline handles the rest. See the Assessment Guide.
Video
Video Analysis
Supported by: Gemini (native multi-modal video understanding)
import { Tekimax, GeminiProvider } from 'tekimax-omat';
const client = new Tekimax({
provider: new GeminiProvider({ apiKey: process.env.GOOGLE_API_KEY! })
});
const analysis = await client.videos.analyze({
video: 'https://cdn.example.org/session-recording.mp4',
model: 'gemini-1.5-flash', // 1M context window — preferred for video
prompt: 'Describe the key moments in this training session recording.',
});
console.log(analysis.content);Note: Video generation is not currently implemented. The
client.videos.generate()method will throw a capability error at runtime. This capability is tracked for a future release.