OEM
2025
Beta version 1.0
DTP/28
VexaBot.AI
Enterprise AI Assistant, Conversational AI, Private LLM, Retrieval-Augmented Generation (RAG), Voice & Chat Automation
Conversational AI for enterprises. Chat and Alexa-style voice, running on a local LLM for privacy with Unified RAG across files, databases, and websites.
Founded to make AI access to enterprise data simple “talk to your data.”
Differentiators
- Privacy- first local LLM option (on-prem/VPC).
- Voice + chat out of the box (Alexa?style interactions).
- Unified RAG: one question ? multi-source retrieval (files/DB/web).
- “DB via MCP” ready: connect through a standard MCP database server (read-only, allowlisted).
- Consulting & enablement: packaged training for leaders, PMs, and engineers.
Solution
- Unified conversational assistant for internal teams & customer?facing flows (chat + voice).
- Unified RAG over enterprise content (files, DBs, selected websites) to answer complex questions with source citations.
- Custom services & AI training to tailor workflows and adoption.
Features
- Chat + Voice assistant UI (web/mobile/telephony via sister brand VexaVoice).
- Knowledge Bases: Files (PDF, docs), Databases (e.g., MySQL/MS-SQL), Web (site scraping/ingestion).
- Deep Research: one-shot plan ? sub-queries (file RAG, DB lookup, web search) ? synthesized answer with citations.
- Admin & Ops: Source mapping, access controls, logs (implementation-dependent).
- Training & Adoption: Enterprise courses and hands?on enablement.
Licenses
Commercial / Proprietary (pricing via sales). Uses open-source connectors where applicable
Review by experts
- Levers: Privacy-first local LLM; voice + chat; multi-source RAG; services & training to accelerate adoption.
- Gaps (public): Limited published specs, customer references, and formal compliance attestations (e.g., SOC 2/HIPAA) on the site confirm during diligence.
- Mitigations: Pilot in a sandbox; enforce strict MCP DB guardrails; require security questionnaire.
Client end Requirment
- Files: Access to approved repositories (share/drive/URLs).
- Databases: Read-only user; schema allowlist; MCP DB server endpoint and credentials.
- Web: Allowed domains for crawling/ingestion.
- Voice: Microphone/telephony setup; optional SIP integration.
- Infra: On-prem or VPC; GPU/CPU sizing per model choice.
Architechture
- LLM layer: Local/self-hosted model where required (privacy).
- Retriever layer: Embedding index + connectors for Files / DB / Web.
- Tooling layer: MCP DB server (external) exposes safe DB tools.
- Orchestrator: Plans multi-source “deep research,” merges results, and returns cited answers.
- Interfaces: Chat UI, Voice (via VexaVoice), APIs for integration.
Infrastructure/Operation
- Setup: Map knowledge sources (files, DB, websites); configure roles/allowlists; index content.
- Run: Users ask in chat/voice ? system plans, retrieves, synthesizes, and cites.
- Admin: Monitor logs, update sources, manage access, retrain embeddings as needed.
Technical Specifications
- DB support (via knowledge base / MCP): MySQL, MS-SQL; broader RDBMS via MCP server.
- Modalities: Text chat; voice I/O.
- Deployment: On-prem or private cloud; customer-controlled data boundary.
- Security: Read-only DB roles, table allowlists, timeouts/row limits (recommended guardrails).
- Languages: Multilingual assistance (marketing claims; confirm locales).
- Reporting: Source-cited answers; activity/logging per deployment.
Current Market
Banking/Healthcare/Govt
Target Clients
- Regulated & data-sensitive sectors: Banking, Healthcare, Government.
- Mid-market/enterprise teams wanting private, multi-source AI assistants.
Pricing / commercial model
Teams & Enterprise plans.
Custom services and training offered.
Use cases
- Banking: Customer support automation; internal policy/contract Q&A; KYC/AML procedure lookup.
- Healthcare: Knowledge lookup, SOP navigation (with PHI kept local).
- Government: Citizen-facing FAQs; internal briefings; document search.
- Sales/Service: Voice agents (VexaVoice) for inbound/outbound workflows.
Differentiators
- Privacy- first local LLM option (on-prem/VPC).
- Voice + chat out of the box (Alexa?style interactions).
- Unified RAG: one question ? multi-source retrieval (files/DB/web).
- “DB via MCP” ready: connect through a standard MCP database server (read-only, allowlisted).
- Consulting & enablement: packaged training for leaders, PMs, and engineers.
Solution
- Unified conversational assistant for internal teams & customer?facing flows (chat + voice).
- Unified RAG over enterprise content (files, DBs, selected websites) to answer complex questions with source citations.
- Custom services & AI training to tailor workflows and adoption.
Features
- Chat + Voice assistant UI (web/mobile/telephony via sister brand VexaVoice).
- Knowledge Bases: Files (PDF, docs), Databases (e.g., MySQL/MS-SQL), Web (site scraping/ingestion).
- Deep Research: one-shot plan ? sub-queries (file RAG, DB lookup, web search) ? synthesized answer with citations.
- Admin & Ops: Source mapping, access controls, logs (implementation-dependent).
- Training & Adoption: Enterprise courses and hands?on enablement.
Licenses
Commercial / Proprietary (pricing via sales). Uses open-source connectors where applicable
Review by experts
- Levers: Privacy-first local LLM; voice + chat; multi-source RAG; services & training to accelerate adoption.
- Gaps (public): Limited published specs, customer references, and formal compliance attestations (e.g., SOC 2/HIPAA) on the site confirm during diligence.
- Mitigations: Pilot in a sandbox; enforce strict MCP DB guardrails; require security questionnaire.
Client end Requirment
- Files: Access to approved repositories (share/drive/URLs).
- Databases: Read-only user; schema allowlist; MCP DB server endpoint and credentials.
- Web: Allowed domains for crawling/ingestion.
- Voice: Microphone/telephony setup; optional SIP integration.
- Infra: On-prem or VPC; GPU/CPU sizing per model choice.
Scope
Architechture
- LLM layer: Local/self-hosted model where required (privacy).
- Retriever layer: Embedding index + connectors for Files / DB / Web.
- Tooling layer: MCP DB server (external) exposes safe DB tools.
- Orchestrator: Plans multi-source “deep research,” merges results, and returns cited answers.
- Interfaces: Chat UI, Voice (via VexaVoice), APIs for integration.
Infrastructure/Operation
- Setup: Map knowledge sources (files, DB, websites); configure roles/allowlists; index content.
- Run: Users ask in chat/voice ? system plans, retrieves, synthesizes, and cites.
- Admin: Monitor logs, update sources, manage access, retrain embeddings as needed.
Technical Specifications
- DB support (via knowledge base / MCP): MySQL, MS-SQL; broader RDBMS via MCP server.
- Modalities: Text chat; voice I/O.
- Deployment: On-prem or private cloud; customer-controlled data boundary.
- Security: Read-only DB roles, table allowlists, timeouts/row limits (recommended guardrails).
- Languages: Multilingual assistance (marketing claims; confirm locales).
- Reporting: Source-cited answers; activity/logging per deployment.
Current Market
Banking/Healthcare/Govt
Target Clients
- Regulated & data-sensitive sectors: Banking, Healthcare, Government.
- Mid-market/enterprise teams wanting private, multi-source AI assistants.
Pricing / commercial model
Teams & Enterprise plans.
Custom services and training offered.
Use cases
- Banking: Customer support automation; internal policy/contract Q&A; KYC/AML procedure lookup.
- Healthcare: Knowledge lookup, SOP navigation (with PHI kept local).
- Government: Citizen-facing FAQs; internal briefings; document search.
- Sales/Service: Voice agents (VexaVoice) for inbound/outbound workflows.
