{"id":"weberg619-cadre-ai","name":"cadre-ai","homepage":"https://www.bimopsstudio.com","repo_url":"https://github.com/WeberG619/cadre-ai","category":"ai-ml","subcategories":[],"tags":["ai-agents","voice","bim","revit","mcp","fastapi","websockets","cloud-run"],"what_it_does":"Cadre-AI is a voice-controlled AI agent that streams audio via WebSockets to a FastAPI server, uses Gemini Live for real-time speech and tool orchestration, and executes actions through MCP toolsets (financial data, web search/weather, and optionally Revit integration via a named-pipe proxy). It supports a local mode (including Revit when enabled) and a Cloud Run mode (Revit disabled).","use_cases":["Voice-driven BIM queries (e.g., levels, rooms, walls, doors/windows, sheets)","Voice-driven BIM edits/creation (e.g., create walls/doors/windows/rooms, place views on sheets)","QA/QC and compliance checks on building models via tool execution","Financial analysis and portfolio/risk assistance via MCP toolsets and market/news data","Voice search and weather/code/material lookup for building-related planning"],"not_for":["Hard real-time safety-critical control where human confirmation is required","Production deployments needing a well-defined public API contract with strong backward-compat guarantees","Environments that require strict compliance certifications where the security posture of the deployment is not documented"],"best_when":"You want a prototype-to-pilot voice agent that can orchestrate BIM + financial + web tools, especially where local Revit integration is acceptable and you can control the deployment environment.","avoid_when":"You need strict rate-limit guarantees, strong operational reliability documentation (SLA, retry/idempotency guarantees), or you cannot safely manage access to a tool-executing voice server that can potentially trigger model modifications.","alternatives":["Custom voice-to-tool orchestration using an agent framework (e.g., LangChain/LangGraph) with your own tool APIs/MCP servers","Revit automation via scripting/RPA (e.g., pyRevit/BIM automation scripts) triggered by a separate voice layer","General-purpose MCP-capable agent shells with task-specific toolsets for BIM/finance/web"],"af_score":39.5,"security_score":46.8,"reliability_score":22.5,"package_type":"mcp_server","discovery_source":["github"],"priority":"high","status":"evaluated","version_evaluated":null,"last_evaluated":"2026-03-30T15:41:57.965218+00:00","interface":{"has_rest_api":false,"has_graphql":false,"has_grpc":false,"has_mcp_server":true,"mcp_server_url":null,"has_sdk":false,"sdk_languages":[],"openapi_spec_url":null,"webhooks":false},"auth":{"methods":["Google API key via GOOGLE_API_KEY for Gemini","Cloud Run IAM / allow-unauthenticated indicates deploy-time access control (per README command)"],"oauth":false,"scopes":false,"notes":"Auth for tool execution and the WebSocket server is not clearly described in the README. Cloud Run example uses --allow-unauthenticated, which may expose the service unless constrained by network/IAM elsewhere."},"pricing":{"model":"gemini-2.5-flash-native-audio-latest (voice) / gem","free_tier_exists":false,"free_tier_limits":null,"paid_tiers":[],"requires_credit_card":false,"estimated_workload_costs":null,"notes":"Pricing is not described in the README; costs likely depend on Gemini API usage and any upstream data providers (Finnhub/Alpha Vantage/FRED/Google CSE), if enabled."},"requirements":{"requires_signup":false,"requires_credit_card":false,"domain_verification":false,"data_residency":[],"compliance":[],"min_contract":null},"agent_readiness":{"af_score":39.5,"security_score":46.8,"reliability_score":22.5,"mcp_server_quality":55.0,"documentation_accuracy":45.0,"error_message_quality":0.0,"error_message_notes":null,"auth_complexity":55.0,"rate_limit_clarity":10.0,"tls_enforcement":85.0,"auth_strength":35.0,"scope_granularity":10.0,"dependency_hygiene":40.0,"secret_handling":65.0,"security_notes":"README indicates HTTPS usage for local UI (self-signed certs) and a Cloud Run deployment, but does not describe authentication/authorization for the WebSocket endpoint or tool execution. Service access may be broad if Cloud Run is deployed with --allow-unauthenticated. Secrets are provided via environment variables, but there is no documented guarantee that logs never include them. Tool execution (including Revit modifications when enabled) increases the impact of misuse; guardrails and least-privilege controls are not documented.","uptime_documented":0.0,"version_stability":35.0,"breaking_changes_history":30.0,"error_recovery":25.0,"idempotency_support":"false","idempotency_notes":"README describes tool execution but does not document idempotency semantics for BIM-modifying actions or how repeated tool calls are handled.","pagination_style":"none","retry_guidance_documented":false,"known_agent_gotchas":["Voice-to-tool systems may execute irreversible BIM actions; ensure you add confirmation/guardrails at the agent/tool layer.","Revit integration is disabled in cloud mode (REVIT_ENABLED=false), so tool availability differs by deployment.","MCP tools are executed via stdio subprocesses; orchestration can be brittle without clear timeout/crash recovery behavior."]}}