OrionSoft's enterprise conversational agent sends every question down the safest path, answers with citations, calculates exactly, and turns website visitors into booked discovery calls.
"Which chatbot services does OrionSoft offer?"
"Budget is $184,500. Cut spend by 17%. What is the net budget?"
"Do you offer a 40% lifetime discount?"
Moving from prototype to a system customers can rely on means solving five problems at the same time.
When documents are unclear, language models fill the gaps. One wrong answer on pricing or policy becomes a liability.
An evidence check decides if the material is strong enough, and the agent says so when it is not.
An agent with web search can send confidential terms, such as salaries or budgets, to third-party search APIs.
An anti-leak gate keeps confidential queries away from external tools.
Semantic search misses exact codes and acronyms. Keyword search misses meaning and paraphrasing.
Hybrid retrieval uses both, then a reranker removes false positives.
Models return numbers that sound right but are wrong, especially with percentages and multi-tier discounts.
A deterministic engine solves the maths, so the model never guesses.
Contact forms attract spam and get abandoned, and ordinary chat widgets rarely verify who they talk to.
Visitors verify themselves in chat and book a call in one click.
An autonomous router classifies each question into one of five execution paths. Pick a path to see what it does.
Sends the question to the hybrid dense and keyword retrieval pipeline. internal_rag
Searches the connected Google Drive workspace. google_drive
Solves maths with a deterministic engine built on Python Abstract Syntax Trees. calculator
Looks up external technology releases. web_search
Politely refuses questions that are unanswerable or out of scope. abstain
The anti-exfiltration gate sits in front of every path. Any query that touches internal employee policies, compensation data or other confidential keywords is barred from reaching external search tools. That closes the door on prompt-injection attacks designed to leak private information.
No single search method is good enough, so every document is indexed two ways and the results are cleaned up before the model writes a word.
A dense vector index (384 dimensions) captures meaning and context.
A BM25 keyword index catches acronyms, product codes and identifiers.
The top results from both searches are combined.
A cross-encoder reads the question and each chunk together and removes false positives.
Is the material strong enough to answer from?
Only then does the model write a reply.
Two guardrails that stop a confident-sounding wrong answer from ever reaching a customer.
"If our cloud budget is $184,500 and we reduce spend by 17%, what is the net budget?"
An evidence evaluator scores the retrieved chunks on three signals before any answer is written.
The agent answers, with citations.
The agent replies that the available company documents do not specify this information.
The same widget handles support, sales and scheduling, so visitors do not have to leave the chat or fill in a form.
The visitor shares their name, company and email. Syntax, disposable domains (30+ providers blocked) and live DNS are checked, and a one-time code confirms the address.
Qualified inquiries go straight to the sales leadership inbox with an AI conversation summary.
When the chat shows buying intent, the agent saves the record and creates links for Google Calendar and Outlook or Office 365, plus a standard .ics invite.
The pipeline was tested with automated suites: 15 real-world scenario prompts for routing, and 29 domain queries for retrieval.
100% (15/15)
90% or higher
100%
95% or higher
100%
90% or higher
0.0%
5% or lower
100%
100%
100%
100%
100%
95% or higher
1,137.14 ms
1,500 ms or lower
100%
The correct source appeared in the top 3 for every query.
1.0000
The most relevant chunk ranked first across all 29 queries.
100%
No irrelevant context was pushed into the model's prompt.
100%
All the ground-truth evidence was captured.
Good architecture is a series of choices with a cost. These are the three biggest, and why each was worth it.
The embedding and reranker models inside the Docker image add about 1.2 GB to the build.
No third-party rate limits, retrieval under 150 ms, no external data transmission and reproducible offline results.
It needs dedicated parsing grammars for natural-language maths expressions.
Zero arithmetic errors, hundreds of milliseconds saved per question and no code-injection surface.
It adds some write-coordination logic across local and cloud storage.
If the cloud connection drops, leads, chats and bookings keep saving locally, with no data loss.
The widget runs in an isolated iframe, so it never clashes with your site's stylesheets.
<script src="https://your-domain.com/widget-loader.js" async></script>
Book a technical discovery call and see the router, retrieval and guardrails on your own documents.