Roadmap
| Quarter | Feature | Status | Description |
|---|---|---|---|
| Q4 2023 | FastAPI Refactor |  |
Refactored to an all-components-in-one-repo codebase |
| LLM-powered search | |
Match user questions to content in the database using embeddings from LLMs (Open AI) | |
| Q1 2024 | LLM responses | |
Craft a custom response to the question using LLM chat and the content in the database |
| AI Guardrails | |
Keep LLM responses friendly and strictly context-based | |
| Multi-lingual content |  |
Support storage and matching of multi-language variants of FAQs | |
| Q2 2024 | Support for low resource language |  |
Support question-answering in regional languages. First targets: Xhosa, Zulu, Hindi, Igbo, Hausa |
| Intent Classification & Message Triaging | |
Tag user messages with intents & flag urgency | |
| Admin app 2.0 | |
A new and better look for Admin App that supports multilingual content management | |
| Q3 2024 | Content Management Dashboard & Engineering Dashboard |  |
Monitor uptime, response rates, throughput HTTP response codes. See content use, questions that receive poor feedback, missing content, and more. |
| Multi-turn chat | |
Refine or clarify user question through conversation. | |
| Document-based question answering | |
Retrieve answers from a document set | |
| Q4 2024 | Voice notes support | |
Automatic Speech Recognition for audio message to content matching |
| Multimedia content | |
Respond with not just text but images and audio as well. | |
| A/B Testing | |
Test and decide content that works better for users |
Key
- Completed
- In active development
- Queued - Yet to be scoped
Beyond 2024
- Privately hosted LLM modules
- Multi-tenant architecture
- User configurable Guardrails (in Admin app)
- Answer contextualization by user demographic
- Feedback-based model/LLM fine tuning
- Safety tests, fairness reports and ethical reviews