# portfolio-optimiser — environment template. Copy to `.env` and fill in. NEVER commit `.env`.
#
# Two backend profiles (D2): LOCAL (OpenAI-compatible local endpoint, dev default — D6) and
# AZURE (Foundry, for targeted minimal verification against the private tenant).
#
# Named `env.template` (no leading dot) on purpose: a leading-dot `.env.template` is blocked by
# the llm-security pre-write pathguard. Copy to `.env` locally (`.env` is gitignored).

# --- LOCAL profile (dev default) — OpenAI-compatible local endpoint (Ollama / LM Studio) ---
# No-egress notes (research 03): the endpoint MUST be loopback; a model pull is the ONLY
# legitimate egress and is explicit. Harden Ollama: bind to 127.0.0.1, pin >= 0.17.1, and
# leave OLLAMA_DEBUG unset.
PORTFOLIO_LOCAL_BASE_URL=http://127.0.0.1:11434/v1   # loopback only — never a remote host
PORTFOLIO_LOCAL_API_KEY=ollama                       # local servers accept any non-empty key
# Recommended Ollama hardening (set in the Ollama service environment, not here):
#   OLLAMA_HOST=127.0.0.1:11434   # bind loopback only
#   (pin Ollama >= 0.17.1; leave OLLAMA_DEBUG unset)

# --- AZURE / Foundry profile (targeted verification only — D6) ---
# Deployment (model) names are tenant-specific and live in data/model_map.json (azure.*).
# The credential is chosen by the ENVIRONMENT and is not configured here:
#   * developer host        -> AzureCliCredential; constructing it acquires no token, so `az login`
#                              stays your own manual step (docs/2026-07-15-foundry-auth-recipe.md).
#   * Foundry hosted agent  -> ManagedIdentityCredential, selected when FOUNDRY_HOSTING_ENVIRONMENT
#                              holds a non-empty value. A container has no Azure CLI; the platform
#                              mints the agent a dedicated Entra identity at deploy time instead.
PORTFOLIO_FOUNDRY_PROJECT_ENDPOINT=   # tenant-specific Foundry project endpoint (operator supplies)
# Hosted only: Foundry injects FOUNDRY_PROJECT_ENDPOINT itself, and the framework reads it as a
# fallback. The name above wins whenever it holds a non-empty value, so an explicit export is never
# shadowed by a platform default — and an exported-but-empty name falls through rather than masking.

# --- Tracing (U14) — OFF unless this variable holds a value ---
# Unset or empty: OpenTelemetry providers are NOT configured at all. Spans are still made (MAF's
# ENABLE_INSTRUMENTATION defaults to true) and simply discarded, so nothing can leave the process.
#   console  -> spans are written to STDERR. stdout is byte-identical to an untraced run, which is
#               what keeps the pinned demo transcript usable. Refused if any OTEL_EXPORTER_OTLP_*
#               endpoint is also set: those exporters are built unconditionally, so "console"
#               would otherwise be a false statement about where the run's contents went.
#   otlp     -> spans are EXPORTED OVER THE NETWORK, and only to an endpoint you named below.
#               Requires an OTLP exporter package, which this framework deliberately does NOT
#               declare (`pip install opentelemetry-exporter-otlp-proto-grpc`).
# PORTFOLIO_OTEL=console
# OTEL_EXPORTER_OTLP_ENDPOINT=          # required by PORTFOLIO_OTEL=otlp; egress — name it yourself