tekimax-omat is an open-source, Apache-2.0 licensed AI SDK built for organizations doing work that matters — nonprofits, civic tech teams, educational institutions, workforce development programs, healthcare organizations, public-sector teams, and any group building AI tools that serve people rather than extract from them.
It provides a unified, type-safe interface across all major AI providers, a hardened supply chain, and a built-in assessment layer (OMAT) for measuring, auditing, and grounding AI-generated feedback in human-centered evaluation frameworks.
Why tekimax-omat?
- Open by default — Apache-2.0 license. No vendor lock-in, no proprietary SDKs to depend on. Fork it, extend it, publish your own plugins.
- Provider-agnostic — Switch between OpenAI, Anthropic, Gemini, Grok, OpenRouter, and Ollama with a single line change. Your application logic never changes.
- Assessment-ready (OMAT) — Built-in rubric-grounded formative assessment pipeline with fairness auditing, demographic disparity detection, and benchmarking. Built for education (including K–12), workforce training, healthcare, nonprofit program evaluation, and any context where feedback quality and equity matter.
- Security-first — Chainguard-based build artifacts, Cosign signing, Trivy scanning, strict TypeScript, Zod validation at every boundary.
- No vendor bloat — Minimal dependency footprint. Core runtime requires only
zodandeventsource-parser.
Who uses it?
tekimax-omat is designed for any organization building AI tools that serve the public good:
| Sector | What the SDK provides today |
|---|---|
| Education & EdTech | Full OMAT assessment pipeline — rubric-grounded formative feedback, ELL/IEP/FRL fairness auditing, learning progressions, useAssessment React hook |
| Nonprofits | FairnessReport with demographic disparity flags, AIActionTagPlugin audit trails, FAIRMetadata with DOI support for grant-reportable outcomes |
| Civic tech | ApiSkillPlugin to connect any government or community database as an AI-callable tool; PIIFilterPlugin to protect citizen data |
| Research institutions | BenchmarkRunner with Cohen's Kappa accuracy, fairness, actionability, and alignment metrics; FAIR-tagged output with DOI field for archival |
| Workforce development | WorkforceAssessmentPipeline, WorkforceFairnessAuditPlugin, WorkforceDemographicTag (employment status, program, credential target, barriers, O*NET codes), createWorkforceRubric() starter factory |
| Healthcare | HealthLiteracyPipeline, PatientResponse, ClinicalPIIFilterPlugin (MRN, NPI, DEA, ICD-10, NDC, DOB, member ID redaction), pre-built HEALTH_LITERACY_COMPREHENSION_RUBRIC and MEDICATION_INSTRUCTIONS_RUBRIC |
Supported Providers
OpenAI
GPT-4o, o1, DALL-E, Whisper, TTS
Anthropic
Claude 3.5 Sonnet, Claude 3 Opus
Google Gemini
Gemini 1.5 Pro/Flash, video analysis
Ollama
Local models, privacy-preserving
xAI (Grok)
Grok-2, real-time reasoning
OpenRouter
100+ models, single API key
Installation
npm install tekimax-omatQuick Start
import { Tekimax, OpenAIProvider } from 'tekimax-omat';
const provider = new OpenAIProvider({
apiKey: process.env.OPENAI_API_KEY!,
});
const client = new Tekimax({ provider });
const response = await client.text.chat.completions.create({
model: 'gpt-4o',
messages: [{ role: 'user', content: 'Hello!' }],
});
console.log(response.message.content);