Loading Project Media
(Overview)
Anenterprise-ready,local-firstbackendAIassistantconnectingnatural-languagequestionsandspeechinterfacestoModelContextProtocol(MCP)tools.Developedforstructuredinteractionswithexternaltoolsandservicesviaastandardizedprotocollayer.
(Architecture)
BuiltwithTypeScriptandExpressexposingacompliant/mcpendpointoverStreamableHTTPtransportwithServer-SentEvents(SSE).Itfeatures16typedMCPtoolscoveringfraudinvestigation,incidentsimulation,systemdiagnostics,andpersistentmemoryvalidatedthroughZodschemas.
(Engineering)
ThebackendorchestratesspeechprocessingwithWhisper.cppandstreamingaudioAPIs.Executioncontextsenforcestrictdeadlinebudgets,requesttimeouts,andstructurederrorhandlerstomaintainhightoolinvocationreliabilityacrossdesktopandvoiceclients.
- 01
EngineeredtheStreamableHTTPtransportendpointsupportingJSON-RPC2.0andServer-SentEvents.
- 02
Implemented16typedMCPtoolswithZodschemavalidationforautomatedparameterverification.
- 03
Integratedlocalspeech-to-textpipelineswithWhisper.cppandaudiostreamingworkflows.
- 04
Configuredexecutioncontextswithdeadlinebudgetsandgracefultimeouthandling.
Built with
PythonModel Context ProtocolMCP SDKFastAPIExpressTypeScriptZodWhisper.cpp
Explore this project
Inspect the live deployment or browse the repository code.
Haveaprojectinmind?
praveenlk41@gmail.com