EVA Brain v2

EVA Ecosystem Integration

Tool	Purpose	How to Use
37-data-model	Single source of truth for all project entities	GET http://localhost:8010/model/projects/33-eva-brain-v2
29-foundry	Agentic capabilities (search, RAG, eval, observability)	C:\eva-foundry\eva-foundation\29-foundry
48-eva-veritas	Trust score and coverage audit	MCP tool: audit_repo / get_trust_score
07-foundation-layer	Copilot instructions primer + governance templates	MCP tool: apply_primer / audit_project

Agent rule: Query the data model API before reading source files.

Invoke-RestMethod "http://localhost:8010/model/agent-guide"   # complete protocol
Invoke-RestMethod "http://localhost:8010/model/agent-summary" # all layer counts

Enterprise Virtual Assistant - AI Brain Service
Domain: Jurisprudence, case law search, document ingestion
Stack: Python 3.10+, FastAPI, Azure OpenAI, Azure AI Search, Azure Cosmos DB
Status: Code complete ? Sprint 6 Deployment in progress | see STATUS.md
Last Updated: February 25, 2026 @ 09:00 ET

Architecture

EVA Brain v2 is composed of two FastAPI microservices:

Service	Port	Purpose
`eva-brain-api`	8001	Chat, RAG, document ingestion, sessions, assistants
`eva-roles-api`	8002	RBAC resolution, persona switching, cost tag evaluation

All protected eva-brain-api routes call eva-roles-api to validate the actor session and resolve features. Both services are fully present on disk (Sprint 0, Feb 19 ? see STATUS.md).

Azure Services Used

Service	Resource	Purpose
Azure OpenAI	`marco-sandbox-openai`	Chat completions, embeddings
Azure AI Foundry	`marco-sandbox-foundry`	Agent framework, model deployments (`gpt-4o`, `gpt-4o-mini`, `text-embedding-ada-002`)
Azure AI Search	`marco-sandbox-search`	Vector, keyword and hybrid search over jurisprudence indexes
Azure Cosmos DB	`marco-sandbox-cosmos`	Sessions, jobs, audit logs, documents
Azure Blob Storage	`marcosand20260203`	Document upload (raw), processed content, audit trail
Azure Storage Queues	`marcosand20260203`	9-queue pipeline for document processing
Azure Document Intelligence	`marco-sandbox-docint`	PDF parsing, layout and table extraction
Azure Computer Vision	`marco-sandbox-aisvc`	Image OCR and captioning
Azure Text Analytics	(sandbox cognitive)	Entity extraction, PII detection
Azure Translator	(sandbox cognitive)	EN/FR translation
Azure API Management	`marco-sandbox-apim`	Rate limiting, product tiers, auth gateway
Azure Monitor / App Insights	`marco-sandbox-appinsights`	Metrics, traces, logs (OpenTelemetry)

Key Patterns

Request Headers (all protected routes)

X-Actor-OID          -- AAD Object ID of the acting user
X-Correlation-ID     -- Request tracing ID (propagated through all services)
X-Acting-Session     -- Validated session token from eva-roles-api
X-Caller-App         -- Calling application identifier (default: "unknown")
X-Environment        -- Deployment environment (default: "dev")

RBAC

Every protected endpoint is decorated with @require_feature(FeatureID.*). The decorator calls eva-roles-api to validate the session and check feature access. A 403 is returned if the persona does not have the required feature. Public endpoints (config, health, translations) bypass RBAC.

Lane A Logging (no raw content)

No prompts, responses, or document text are ever stored in logs or audit records. Only hashes (SHA-256, 16-char), counts, scores, durations, and IDs are logged. Use hash_content() and create_safe_result_summary() from app.models.agent.

Cost Attribution

Every agent invocation includes CostTags (phase, task, project, costcenter, persona_id, actor_oid). These are used for FinOps charge-back and Azure cost management.

Personas

Defined in services/eva-brain-api/config/personas.yml:

Persona	Tools	Indexes	Max tokens/turn	Autonomy
`legal-researcher`	hybrid_search, semantic_search, keyword_search, chat_completion, log_audit, check_quality	case-law, statutes	4,000	high
`legal-clerk`	keyword_search, chat_completion	case-law	1,000	low
`admin`	all 7 tools	case-law, statutes, policies	8,000	high

API Overview

Group	Endpoints	Status
Chat & RAG	POST /v1/chat, /v1/chat/rrr, /v1/chat/rtr, /v1/chat/work, /v1/chat/hybrid, /v1/chat/ungrounded, /v1/chat/read-retrieve-read, GET /v1/chat/approaches	? Route present
Retrieve	POST /v1/retrieve	? Route present
Sessions	POST/GET/PATCH/DELETE /v1/sessions, /v1/sessions/{id}	? Route present
Documents	GET/PATCH/DELETE /v1/documents/*, /v1/documents/{id}/chunks	download
Ingest	POST /v1/ingest/upload	url, GET /v1/ingest/status/{id}
Tags	GET /v1/tags, /v1/tags/{tag}, /v1/tags/{tag}/documents	? Route present
Search	Saved searches and history	? Route present
Admin	POST /v1/users, GET /v1/groups	? Route present
Config	GET /v1/config/info	features
Health	GET /v1/health, /v1/health/background	? Route present
RBAC (roles)	POST /v1/roles/acting-as, GET /v1/roles/context	features
Assistants	POST /v1/assistant/math/, /v1/assistant/tabular/	? Implemented (Sprint 5)
Logs	GET /v1/logs/audit/, /v1/logs/content/	? Implemented (Sprint 5)
Translations (write)	PUT/PATCH/DELETE /v1/config/translations/{lang}	? Implemented (Sprint 5)
Settings	GET/PATCH /v1/settings	? Implemented (Sprint 5)
Apps Registry	GET/POST/PATCH/DELETE /v1/apps, /v1/apps/{id}/disable	export
Scrum (partial)	GET /v1/scrum/dashboard, /v1/scrum/summary	[CODED] on disk (Sprint 7, untested)
Projects Plane	GET /v1/projects, /v1/projects/{id	blocked
WBS / Critical Path	GET /v1/wbs, /v1/wbs/{id	critical-path
ADO Sync (pull)	POST /v1/ado/sync, GET /v1/ado/sync/status	[PLANNED] Sprint 7
ADO Sync (push)	PATCH /v1/ado/wi/{tag}/state, POST /v1/ado/wi	[PLANNED] Sprint 8
Evidence Proxy	GET /v1/evidence/{id	by-wi

Implemented: 60/60. Planned Sprint 7: +13. Planned Sprint 8: +12.

Document Ingestion Pipeline

Queue-based pipeline using 9 Azure Storage Queues, 11-step state machine:

Upload -> Route -> Parse -> Enrich Text -> Enrich Images -> Chunk -> Embed -> Index -> Update Status -> Log -> Notify

5 document processors: PDF (Azure Document Intelligence), DOCX, XLSX, TXT, Image (Azure Computer Vision)
3 enrichment services: Translation (EN/FR), Entity Extraction, PII Redaction
3 chunking strategies: semantic, fixed (750 tokens + 50 overlap), paragraph
Batch embedding: 16 texts per Azure OpenAI request
Performance target: <2 min upload-to-index for 10-page PDF

All pipeline service files are present on disk. The 6 ingest route endpoints are present in app/routes/ingest.py.

Agent Framework

The agent framework is in app/agents/. Three files are on disk:

conversation_agent.py -- ConversationAgent.invoke(), multi-turn orchestration, 8-hour TTL
tools.py -- AgentTools, 7 registered tools (hybrid_search, semantic_search, keyword_search, chat_completion, log_audit, check_quality, get_rbac_context)
persona_constraints.py -- PersonaConstraintsService, loads from config/personas.yml

Agent models are in app/models/agent.py (540 lines, fully present): ToolInvocation, ThoughtTrace, ReasoningTrace, ConversationTurn, ConversationContext, PersonaConstraints, AgentResponse

The Azure AI Foundry client is in app/clients/foundry_client.py.

Development Setup

Environment Variables

Copy and populate all 20+ variables. Required variables (see tests/conftest.py for full list):

# Azure OpenAI
AZURE_OPENAI_ENDPOINT=https://marco-sandbox-foundry.cognitiveservices.azure.com/
AZURE_OPENAI_DEPLOYMENT=gpt-4o
AZURE_OPENAI_API_KEY=...
AZURE_OPENAI_API_VERSION=2024-02-01

# Azure AI Search
AZURE_SEARCH_ENDPOINT=https://marco-sandbox-search.search.windows.net
AZURE_SEARCH_INDEX=case-law
AZURE_SEARCH_API_KEY=...

# Azure Cosmos DB
AZURE_COSMOS_ENDPOINT=https://marco-sandbox-cosmos.documents.azure.com:443/
AZURE_COSMOS_KEY=...
AZURE_COSMOS_DATABASE=statusdb

# Azure Storage
AZURE_STORAGE_ACCOUNT_NAME=marcosand20260203
AZURE_STORAGE_ACCOUNT_KEY=...
AZURE_STORAGE_CONTAINER=upload

# Azure Cognitive Services
AZURE_COMPUTER_VISION_ENDPOINT=...
AZURE_COMPUTER_VISION_KEY=...
AZURE_TRANSLATOR_ENDPOINT=...
AZURE_TRANSLATOR_KEY=...
AZURE_TRANSLATOR_REGION=canadacentral
AZURE_TEXT_ANALYTICS_ENDPOINT=...
AZURE_TEXT_ANALYTICS_KEY=...

# App
ROLES_API_URL=http://localhost:8002
ENVIRONMENT=dev
LOG_LEVEL=INFO
APPINSIGHTS_CONNECTION_STRING=...

Running Tests

cd services/eva-brain-api
pip install -r requirements.txt
pytest -v --cov=app tests/

Current state: 577/577 tests passing, 72% coverage.

Project Structure

33-eva-brain-v2/
  .github/
    copilot-instructions.md
  contracts/
    schemas/                      -- JSON Schema contracts (6 files)
  docs/
    agent-framework-guide.md
    api-migration-phase-2.md
    eva-faces-integration-guide.md
    performance-benchmarks.md
    runbooks/
  services/
    eva-brain-api/
      app/
        agents/                   -- ConversationAgent, AgentTools, PersonaConstraintsService
        clients/                  -- FoundryClient (Azure AI Projects SDK)
        decorators/               -- feature_guard.py (@require_feature)
        exceptions.py             -- EVAException hierarchy
        middleware/               -- rbac.py (RBACMiddleware)
        models/
          agent.py                -- All agent Pydantic models
          cost.py                 -- CostTags
          features.py             -- FeatureID enum
        routes/                   -- All 11 route files present
        services/                 -- All service files present
        utils/
      config/
        personas.yml
      openapi/
        agent-schemas.yaml
      tests/                      -- 577 tests, 72% coverage
    eva-roles-api/                -- Full service, port 8002
  translations/
    en.json
    fr.json
  README.md                       -- This file
  STATUS.md                       -- Current implementation status
  PLAN.md                         -- Sprint plan (Sprint 6 active)
  ACCEPTANCE.md                   -- Acceptance criteria and sign-off
  FOUNDRY-DEPLOYMENT-GUIDE.md     -- Sprint 6 deployment reference
  FOUNDRY-QUICK-REFERENCE.md      -- Azure CLI quick reference
  DEPLOY-VIA-CODESPACES.md        -- Deployment via GitHub Codespaces

37-data-model Integration ? Query It First, Every Session

The most expensive mistake a Copilot agent makes in this repo is reading source files to answer questions the model already knows.
Before grep_search, before file_search, before read_file ? check http://localhost:8010/model/{layer}/. That is a 3-line query. The alternative is a 10-turn spiral.

37-data-model (port 8010) is the semantic repository for the entire EVA ecosystem ? a Siebel-style object model with 27 layers, query-over-HTTP via FastAPI, and Siebel audit columns (modified_by, row_version, source_file) on every write. It already catalogs brain-v2's services, personas, feature flags, containers, and endpoints.

What it already knows about brain-v2

Layer	What it tracks	Objects
L0 services	`eva-brain-api` (port 8001), `eva-roles-api` (port 8002)	2
L1 personas	`admin`, `legal-researcher`, `legal-clerk`, `machine-agent`, `auditor`	5
L2 feature_flags	`action.chat`, `action.retrieve`, `action.ingest`, `action.search`, `action.admin.*`	10
L3 containers	`sessions`, `jobs`, `audit_logs`, `documents`, `translations`, `settings`, `apps`	13
L4 endpoints	60 brain-v2 endpoints (52 implemented, rest stub/planned)	60
L5 schemas	Request/response shapes for all 60 endpoints	36
L9 infrastructure	`marco-sandbox-*` Azure resources, APIM, Key Vault secrets	23
L25 projects	`33-eva-brain-v2` project record (maturity, ADO epic, pbi_done/total, depends_on)	1
L26 wbs	WBS deliverable nodes for brain-v2 sprints	n

Note: The eva-brain-api service record was updated Feb 25 (row_version=5): status=running, notes=Sprint 5 complete (577/577 tests, 60/60 endpoints, POC trusted-subsystem verified Feb 24). Commit clean: violations=0, exported=3932.

Integration points ? current and planned

1. Feature flags and persona matrix (eva-roles-api, Sprint 6+)

Current state: app/models/features.py (FeatureID enum) and config/personas.yml are both maintained manually. If they diverge from 37-data-model/model/feature_flags.json, the authoritative source is unclear.

Integration: At eva-roles-api startup, rbac_service.py could call:

# Option: query 37-data-model at startup to build persona ? features map
GET http://localhost:8010/model/personas/   # ? list with feature_flags[]
GET http://localhost:8010/model/feature_flags/  # ? each with personas[]

This makes config/personas.yml and FeatureID the runtime cache and 37-data-model the source of truth, eliminating drift. The model already has the correct mapping.

When: Can be done in Sprint 6 or Sprint 7 ? no new endpoints needed in brain-v2.

2. New RBAC flags for PM Plane API (37-data-model must be updated first)

The three new feature flags that Sprint 7 and Sprint 8 require must be added to 37-data-model before @require_feature will work:

Flag	Sprint	Personas	Write cycle
`action.programme`	7	pm_viewer, admin, auditor	`PUT /model/feature_flags/action.programme` ? verify ? export ? assemble
`action.ado_sync`	7	admin, machine-agent	same
`action.ado_write`	8	admin, machine-agent	same

Block: brain-v2's @require_feature("action.programme") decorator will return 403 for everyone until this record exists in 37-data-model and is synced to eva-roles-api's feature map.

3. PM Plane API ? brain-v2 is the HTTP facade over 37-data-model L25/L26 (Sprint 7)

This is the most critical integration. The /v1/projects, /v1/wbs/*, and /v1/scrum/* endpoints that brain-v2 is building in Sprint 7 are APIM-facing proxies over 37-data-model's L25 and L26 layers. Brain-v2 does not own this data ? it reads from and writes back to 37-data-model.

APIM ? brain-v2 /v1/projects   ?  GET http://localhost:8010/model/projects/
APIM ? brain-v2 /v1/wbs        ?  GET http://localhost:8010/model/wbs/
APIM ? brain-v2 /v1/wbs/critical-path  ?  compute from model/wbs/ + GET /model/graph?...

The new service file project_plane_service.py (Sprint 7) must be a typed client over the model API, not over raw JSON files.

4. ADO sync write-back ? 4-step write cycle (Sprint 7?8)

Every ADO sync write from POST /v1/ado/sync or POST /v1/ado/wi must follow the API-first write protocol (captured in copilot-instructions.md):

1. PUT  /model/work_items/{wi_tag}      # stamps modified_at, modified_by, row_version
2. GET  /model/work_items/{wi_tag}      # verify field round-trips correctly
3. POST /model/admin/export             # materialize audit trail ? model/*.json
4. assemble-model.ps1                   # rebuild eva-model.json

Never write directly to model/*.json files ? doing so bypasses row_version and makes GET /model/admin/audit blind to the change. (This anti-pattern was recorded in 37-data-model/STATUS.md, Feb 22, 2026.)

5. Startup health check ? know if 37-data-model is available (Sprint 7)

Brain-v2's GET /v1/health/background should include a model_api field:

# In app/routes/health.py background health:
model_api_ok = await model_client.ping()  # GET http://localhost:8010/health
return { "status": "ok", "model_api": "ok" if model_api_ok else "unavailable" }

If 37-data-model is down, PM Plane endpoints should return 503 + Retry-After, not 500.

6. Impact analysis before any Cosmos schema change

Before modifying any container field in brain-v2 (e.g. adding a field to sessions, renaming a field in jobs):

# Which endpoints and screens break if I change the jobs container?
Invoke-RestMethod "http://localhost:8010/model/impact/?container=jobs"

This is the 37-data-model equivalent of a FK constraint check ? run it before every schema migration PR.

New service: `model_api_client.py`

Sprint 7 should introduce services/eva-brain-api/app/services/model_api_client.py ? a lightweight httpx.AsyncClient wrapper over port 8010:

class ModelAPIClient:
    base_url: str = "http://localhost:8010"

    async def get_projects(self, maturity: str | None = None, category: str | None = None) -> list[dict]
    async def get_wbs(self, level: str | None = None, project_id: str | None = None) -> list[dict]
    async def get_wbs_critical_path(self) -> list[dict]
    async def get_feature_flags(self) -> list[dict]
    async def put_work_item(self, wi_tag: str, payload: dict) -> dict  # write cycle step 1
    async def export(self) -> None                                      # write cycle step 3
    async def ping(self) -> bool

ROLES_API_URL already follows this pattern; add MODEL_API_URL=http://localhost:8010 as a new env var with the same fallback logic.

Related Projects

31-eva-faces -- Frontend UI components (React/TypeScript) ? consumes brain-v2 API (see Downstream Dependencies below)
29-foundry -- EVA Foundry Library: reusable agent skills, MCP servers, Python tools, Prompty templates ? consumes brain-v2 API as a proxy for PM agents (see Downstream Dependencies below)
63-factory-context-auditor -- Workspace context auditing and repo-briefing

Note: Project 63 is now the Factory Context Auditor, not a higher-level agent orchestration repo. Any future agent-orchestration assets should originate from 29-foundry rather than Project 63.
EVA-Jurisprudence-SecMode-Info-Assistant-v1.2 -- Production EVA JP deployment

Downstream Dependencies

31-eva-faces ? brain-v2

eva-faces is a consumer of brain-v2's REST API. The table below tracks which eva-faces pages depend on which brain-v2 endpoint group, and when brain-v2 will deliver them.

eva-faces Page / Component	brain-v2 Endpoint(s)	Brain Sprint	Status
`TranslationsPage`	`PUT/PATCH/DELETE /v1/config/translations/{lang}`	Sprint 5	? Done (Feb 20)
`SettingsPage`	`GET/PATCH /v1/settings`	Sprint 5	? Done (Feb 20)
`AppsPage`	`GET/POST/PATCH/DELETE /v1/apps`	Sprint 5 WI-6	? Done (Feb 20)
`AuditLogsPage`	`GET /v1/logs/audit/*`	Sprint 5	? Done (Feb 20)
All data-plane pages (any live call)	Both services deployed to Azure + APIM routing	Sprint 6	?? In progress (Feb 22 ?)
`ProjectPortfolioPage` (WI-F1?F6)	`GET /v1/projects`, `GET /v1/projects/{id}`, `GET /v1/projects/blocked`, `GET /v1/projects/by-stream`	Sprint 7	?? Planned
`WBSTreePage` + `CriticalPathPanel` (WI-F10?F18)	`GET /v1/wbs`, `GET /v1/wbs/{id}`, `GET /v1/wbs/critical-path`, `GET /v1/wbs/by-sprint`	Sprint 7	?? Planned
`SprintBoardPage` + `SprintSummaryStrip` + `VelocityChart` (WI-F19?F26)	`GET /v1/scrum/dashboard`, `GET /v1/scrum/summary`, `GET /v1/scrum/velocity`, `GET /v1/scrum/products`	Sprint 7	?? Planned
`DomainAssistantChatPage` (Phase 3)	`POST /v1/chat` (SSE streaming) + `GET/POST/DELETE /v1/sessions` + `GET /v1/chat/approaches`	Sprint 6 (deploy)	?? Planned (live after Sprint 6 Azure deploy)
`EvidenceViewerPanel` (WI-F27+, Epic 4)	`GET /v1/evidence/{id}`, `GET /v1/evidence/by-wi/{tag}`, `GET /v1/runs`, `GET /v1/evidence/summary`	Sprint 8	?? Planned
`RbacPage` / `RbacRolesPage`	`GET /v1/roles/personas`, `POST /v1/roles/acting-as`, `GET /v1/roles/features` (eva-roles-api)	Sprint 1	? Done (Feb 19)

Roadmap for eva-faces unblocking:

Sprint 6  (active)  ? Both services deployed + APIM routing ? all live calls become possible
Sprint 7  (planned) ? PM Plane API live ? ProjectPortfolioPage, WBSTreePage, SprintBoardPage
Sprint 7  (planned) ? Scrum Dashboard + ADO Pull Sync ? SprintBoardPage live data
Sprint 8  (planned) ? Evidence Integration ? EvidenceViewerPanel live

RBAC flags that must be added to 37-data-model before eva-faces can use them: action.programme (Sprint 7), action.ado_sync (Sprint 7), action.ado_write (Sprint 8).

29-foundry ? brain-v2

29-foundry's PM agents use brain-v2 as a data proxy ? all PM Plane API calls from foundry agents go through brain-v2 rather than calling 37-data-model directly. The table below lists foundry's runtime tool calls.

29-foundry Agent / Tool	brain-v2 Endpoint(s)	Brain Sprint	Status
`pm-agent.py` ? tool `get_projects()` (WI-A2)	`GET /v1/projects`	Sprint 7	?? Planned
`pm-agent.py` ? tool `get_wbs()` (WI-A3)	`GET /v1/wbs`	Sprint 7	?? Planned
`pm-agent.py` ? tool `get_critical_path()` (WI-A4)	`GET /v1/wbs/critical-path`	Sprint 7	?? Planned
`pm-agent.py` ? tool `get_sprint_status()` (WI-A5)	`GET /v1/scrum/summary`, `GET /v1/scrum/dashboard`	Sprint 7	?? Planned
`pm-agent.py` ? tool `get_evidence_summary()` (WI-A6)	`GET /v1/evidence/summary`	Sprint 8	?? Planned
`pm-agent-server.py` HTTP wrapper (WI-A9)	Full brain-v2 API on port 8001 (sandbox)	Sprint 6 deploy	?? Planned
`ado-sync-agent.py` ? ADO pull sync (WI-A12)	`POST /v1/ado/sync`, `GET /v1/ado/sync/status`	Sprint 7	?? Planned
`sprint-planner-agent.py` ? tool `create_work_item()` (WI-A22)	`POST /v1/ado/wi`	Sprint 8	?? Planned

Structural (non-runtime) dependency: 29-foundry also imports ConversationAgent and AgentTools from brain-v2 as reference patterns into copilot-skills/azure-plane/eva-brain-v2/ and agent-framework/. This is a one-time code import, not a live call.

Roadmap for 29-foundry unblocking:

Sprint 6  (active)  ? brain-v2 deployed to sandbox ? pm-agent-server.py can point at real endpoint
Sprint 7  (planned) ? PM Plane API (/v1/projects, /v1/wbs, /v1/scrum/*) live ? WI-A2?A6, WI-A9, WI-A12 unblocked
Sprint 8  (planned) ? Evidence + ADO Push (/v1/evidence/*, /v1/ado/wi) live ? WI-A6 (evidence_summary), WI-A22 unblocked

Name		Name	Last commit message	Last commit date
Latest commit History 7 Commits
.eva		.eva
.github		.github
.memories/session		.memories/session
archive/completed-plans		archive/completed-plans
ci		ci
contracts/schemas		contracts/schemas
docs		docs
evidence		evidence
scripts		scripts
services		services
translations		translations
.env.ado.example		.env.ado.example
.env.example		.env.example
.gitignore		.gitignore
ACCEPTANCE.md		ACCEPTANCE.md
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
PLAN.md		PLAN.md
PLAN.md.bak		PLAN.md.bak
README.md		README.md
SECURITY.md		SECURITY.md
STATUS.md		STATUS.md
acr-build.log		acr-build.log
ado-artifacts.json		ado-artifacts.json
ado-import.ps1		ado-import.ps1
ado-setup-output.txt		ado-setup-output.txt
current-template.json		current-template.json

Folders and files

Latest commit

History

Repository files navigation

EVA Brain v2

EVA Ecosystem Integration

Architecture

Azure Services Used

Key Patterns

Request Headers (all protected routes)

RBAC

Lane A Logging (no raw content)

Cost Attribution

Personas

API Overview

Document Ingestion Pipeline

Agent Framework

Development Setup

Environment Variables

Running Tests

Project Structure

37-data-model Integration ? Query It First, Every Session

What it already knows about brain-v2

Integration points ? current and planned

1. Feature flags and persona matrix (eva-roles-api, Sprint 6+)

2. New RBAC flags for PM Plane API (37-data-model must be updated first)

3. PM Plane API ? brain-v2 is the HTTP facade over 37-data-model L25/L26 (Sprint 7)

4. ADO sync write-back ? 4-step write cycle (Sprint 7?8)

5. Startup health check ? know if 37-data-model is available (Sprint 7)

6. Impact analysis before any Cosmos schema change

New service: model_api_client.py

Related Projects

Downstream Dependencies

31-eva-faces ? brain-v2

29-foundry ? brain-v2

About

Topics

Resources

License

Code of conduct

Contributing

Security policy

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

New service: `model_api_client.py`

Packages