Internal knowledge / grounded support search
SUPPORT KNOWLEDGE / GROUNDED SEARCH
01 / 06BriopediaA support library with its source attached.
A versioned knowledge library paired with retrieval and conversational search for product support teams.
Prior Brio team product experience. Not a Teregon client engagement.

THE WORK / SUPPORT KNOWLEDGE
02 / 06A useful answer starts
with maintained material.

Support material changes over time. Agents need to find the relevant guidance and know which source supports an answer.
Briopedia paired an editable, git-backed wiki with semantic retrieval and an LLM chat layer, so source material could be maintained separately from the answer interface.
No authenticated Briopedia interface is reproduced in this presentation.
THE KNOWLEDGE PATH / FROM EDIT TO ANSWER
03 / 06Keep the source.
Find the passage.
The library, retrieval layer, and chat interface each have a clear job.
- 01EDITMaintainProduct knowledge in a git-backed wiki
- 02TRACKVersionReview changes and preserve history
- 03SEARCHRetrieveFAISS search finds relevant passages
- 04DRAFTRespondA local or cloud LLM uses retrieved context
- 05CITEReferenceReturn the answer with its source material
Illustrative workflow; the cards above are not captures of the signed-in application.
ILLUSTRATIVE INTERFACE / SOURCE-GROUNDED ANSWER
04 / 06A clear path from question
back to the source.
KNOWLEDGE
Water dispenser guideFilter replacementTemperature issueService policyProduct guide · updated
Use the approved product guide to identify the unit’s filter procedure. Confirm the model before sharing steps.
Illustrative interface only. Text is generic and is not product guidance or a live Briopedia capture.
TECHNICAL SHAPE / PACKAGED FOR A TEAM
05 / 06A replaceable model layer.
A retained knowledge base.
SUPPORT INTERFACE→FastAPI
APPLICATION API→FAISS
RETRIEVAL INDEX
SourceGit-backed wiki with editable content and reviewable history.
ModelLocal or cloud LLM connection, kept replaceable.
DeploymentInterface, API, and model connection packaged with Docker Compose.
Architecture follows the published case-study description. No performance, revenue, or customer-impact figures are asserted.
TEREGON LABS / APPLIED AI + SOFTWARE
06 / 06Search should lead
back to the source.
The system’s durable part is the maintained knowledge. Retrieval and language models help people reach it; citations and editable history keep the answer connected to its source.