Consolidate runtime secrets

cloudrun/service.yaml

@@ -54,17 +54,12 @@ spec:
               value: "text-embedding-3-small"
             - name: EMBEDDING_DIMENSIONS
               value: "1536"
-            - name: OPENAI_API_KEY
+            - name: PROVIDER_CONFIG_JSON
               valueFrom:
                 secretKeyRef:
-                  name: openai-api-key
+                  name: provider-config
                   key: latest
             # --- Search ---
-            - name: TAVILY_API_KEY
-              valueFrom:
-                secretKeyRef:
-                  name: tavily-api-key
-                  key: latest
             # --- Graph store (Neo4j Aura) ---
             # URI and username are not sensitive; only the password lives in Secret Manager.
             - name: NEO4J_URI
@@ -93,16 +88,28 @@ spec:
             - name: READINESS_REQUIRE_NEO4J
               value: "true"
             # --- Inngest ---
-            - name: INNGEST_EVENT_KEY
+            - name: INNGEST_CONFIG_JSON
               valueFrom:
                 secretKeyRef:
-                  name: inngest-event-key
+                  name: inngest-config
                   key: latest
-            - name: INNGEST_SIGNING_KEY
+            # --- Billing (Stripe) ---
+            - name: BILLING_CONFIG_JSON
               valueFrom:
                 secretKeyRef:
-                  name: inngest-signing-key
+                  name: billing-config
                   key: latest
+            # --- Redis cache ---
+            # REDIS_URL is bundled into PROVIDER_CONFIG_JSON to stay within the
+            # Secret Manager free active-version allowance.
+            - name: REDIS_CACHE_TTL_DB_LIST_SECONDS
+              value: "30"
+            - name: STRIPE_SUCCESS_URL
+              value: "https://cortex-ai-prod.vercel.app/billing/success"
+            - name: STRIPE_CANCEL_URL
+              value: "https://cortex-ai-prod.vercel.app/billing/cancel"
+            - name: STRIPE_PORTAL_RETURN_URL
+              value: "https://cortex-ai-prod.vercel.app"
             # --- CORS ---
             - name: CORS_ORIGINS
               value: '["https://cortex-ai-prod.vercel.app"]'
@@ -113,9 +120,7 @@ spec:
                   name: internal-dispatch-secret
                   key: latest
             # --- Observability ---
-            # ponytail: observability off for the free-tier deploy; re-add
-            # langsmith-api-key / langfuse-* secrets when tracing matters.
             - name: LANGSMITH_TRACING
               value: "false"
             - name: LANGFUSE_ENABLED
-              value: "false"
+              value: "true"

docs/billing.md

@@ -4,7 +4,7 @@ Cortex uses Stripe Hosted Checkout, signed webhooks, and the customer portal for
 
 ## Configuration
 
-Required variables:
+Required local-development variables:
 
 - `STRIPE_SECRET_KEY`
 - `STRIPE_WEBHOOK_SECRET`
@@ -13,6 +13,11 @@ Required variables:
 - `STRIPE_CANCEL_URL`
 - `STRIPE_PORTAL_RETURN_URL`
 
+In production, Cloud Run uses one `BILLING_CONFIG_JSON` Secret Manager value containing
+`stripe_secret_key`, `stripe_webhook_secret`, and `stripe_pro_price_id`; the three redirect
+URLs remain ordinary environment variables. This keeps the Stripe credentials within one
+active secret version.
+
 Store production secret values in Google Secret Manager. See [Production configuration](env-vars-production.md).
 
 ## Webhook setup
@@ -46,4 +51,3 @@ flowchart LR
 ```
 
 Webhook events are the authoritative source for subscription state. The backend verifies the Stripe signature before updating `user_subscriptions`; research and chat quota guards then read the synchronized subscription and usage state.
-

docs/env-vars-production.md

@@ -8,15 +8,14 @@ All secrets should be stored in **Google Secret Manager** and referenced via `va
 
 | Env Var | Secret Manager | Notes |
 |---|---|---|
-| `OPENAI_API_KEY` | yes | Required when `LLM_PROVIDER=openai` |
+| `PROVIDER_CONFIG_JSON` | yes | Secret JSON with provider, Redis, Langfuse, and LangSmith settings: `openai_api_key`, `tavily_api_key`, `redis_url`, `langfuse_public_key`, `langfuse_secret_key`, `langfuse_base_url`, `langsmith_api_key`, `langsmith_project`, `langsmith_endpoint`, `langsmith_redaction_mode`, `langsmith_sampling_rate`, and `langsmith_tracing`. Required when `LLM_PROVIDER=openai`. |
 | `OPENROUTER_API_KEY` | yes | Required when `LLM_PROVIDER=openrouter` |
 | `LLM_PROVIDER` | no | `openai` or `openrouter` |
 | `OPENAI_MODEL` | no | e.g. `gpt-4o-mini` |
 | `ROUTER_TIMEOUT_SECONDS` | no | Maximum duration of each router LLM request; production uses `10.0` |
 | `EMBEDDING_PROVIDER` | no | `openai` |
 | `EMBEDDING_MODEL` | no | `text-embedding-3-small` |
 | `EMBEDDING_DIMENSIONS` | no | `1536` |
-| `TAVILY_API_KEY` | yes | |
 | `ALPHA_VANTAGE_API_KEY` | yes | Required when `ASSET_PRICE_PROVIDER=alphavantage_mcp` |
 | `ASSET_PRICE_PROVIDER` | no | `alphavantage_mcp` |
 | `ALPHA_VANTAGE_MCP_URL` | no | Optional full remote MCP URL override |
@@ -32,12 +31,9 @@ All secrets should be stored in **Google Secret Manager** and referenced via `va
 | `READINESS_REQUIRE_SUPABASE` | no | Set to `true` in production; `/ready` returns `503` if Supabase is missing or unavailable |
 | `READINESS_REQUIRE_NEO4J` | no | Set to `true` in production; `/ready` returns `503` if Neo4j is missing or unavailable |
 | `READINESS_TIMEOUT_SECONDS` | no | Per-dependency probe timeout; defaults to `2.0` seconds |
-| `INNGEST_EVENT_KEY` | yes | From Inngest dashboard → Keys |
-| `INNGEST_SIGNING_KEY` | yes | From Inngest dashboard → Keys. **Must not be empty in prod.** |
-| `REDIS_URL` | yes | Upstash `rediss://` URL |
-| `STRIPE_SECRET_KEY` | yes | |
-| `STRIPE_WEBHOOK_SECRET` | yes | Update Stripe webhook URL to `https://<cloud-run-url>/api/billing/webhook` after first deploy |
-| `STRIPE_PRO_PRICE_ID` | no | |
+| `INNGEST_CONFIG_JSON` | yes | Secret JSON with `inngest_event_key` and `inngest_signing_key` from the Inngest dashboard. |
+| `REDIS_CACHE_TTL_DB_LIST_SECONDS` | no | Cache TTL for DB-heavy list/read-all responses; production uses `30`. `REDIS_URL` is supplied through `PROVIDER_CONFIG_JSON`. |
+| `BILLING_CONFIG_JSON` | yes | Secret JSON with `stripe_secret_key`, `stripe_webhook_secret`, and `stripe_pro_price_id`. This bundles Stripe credentials into one Secret Manager version. |
 | `STRIPE_SUCCESS_URL` | no | `https://<frontend>/billing/success` |
 | `STRIPE_CANCEL_URL` | no | `https://<frontend>/billing/cancel` |
 | `STRIPE_PORTAL_RETURN_URL` | no | `https://<frontend>` |
@@ -49,9 +45,8 @@ All secrets should be stored in **Google Secret Manager** and referenced via `va
 | Env Var | Secret Manager | Notes |
 |---|---|---|
 | `COHERE_API_KEY` | yes | Enables cross-encoder reranking |
-| `LANGFUSE_PUBLIC_KEY` | yes | |
-| `LANGFUSE_SECRET_KEY` | yes | |
-| `LANGFUSE_ENABLED` | no | Set to `true` only when both LangFuse keys are configured; otherwise feedback remains stored locally |
+| `LANGFUSE_ENABLED` | no | Production sets this to `true`; Langfuse credentials and base URL are supplied through `PROVIDER_CONFIG_JSON`. |
+| `LANGSMITH_*` | no | LangSmith tracing, API key, project, endpoint, redaction mode, and sampling rate are supplied through `PROVIDER_CONFIG_JSON`. |
 | `LANGFUSE_ENV` | no | `prod` |
 | `LANGFUSE_RELEASE` | no | Git SHA or semver, e.g. `v1.2.3` |
 | `SENTRY_DSN` | yes | Error tracking; unset disables it entirely |
@@ -91,19 +86,16 @@ PROJECT=cortex-496709
 REGION=your-region
 
 secrets=(
-  openai-api-key
-  tavily-api-key
+  provider-config
   cohere-api-key
   neo4j-uri
   neo4j-username
   neo4j-password
   supabase-secret-key
   supabase-jwt-secret
-  inngest-event-key
-  inngest-signing-key
+  inngest-config
   redis-url
-  stripe-secret-key
-  stripe-webhook-secret
+  billing-config
   internal-dispatch-secret
   langfuse-public-key
   langfuse-secret-key

src/config.py

@@ -1,9 +1,33 @@
 """Application settings loaded from environment variables."""
 
-from pydantic import AliasChoices, Field, field_validator
+import json
+
+from pydantic import AliasChoices, Field, field_validator, model_validator
 from pydantic_settings import BaseSettings, SettingsConfigDict
 
 
+def _parse_bundled_secret(
+    raw: str,
+    config_name: str,
+    required: tuple[str, ...],
+) -> dict[str, str]:
+    try:
+        config = json.loads(raw)
+    except json.JSONDecodeError as exc:
+        raise ValueError(f"{config_name} must be valid JSON.") from exc
+    if not isinstance(config, dict):
+        raise ValueError(f"{config_name} must be a JSON object.")
+
+    values: dict[str, str] = {}
+    for name in required:
+        value = config.get(name)
+        if not isinstance(value, str) or not value:
+            fields = ", ".join(required)
+            raise ValueError(f"{config_name} must include non-empty {fields} values.")
+        values[name] = value
+    return values
+
+
 class Settings(BaseSettings):
     model_config = SettingsConfigDict(
         env_file=".env",
@@ -13,6 +37,10 @@ class Settings(BaseSettings):
 
     # LLM
     llm_provider: str = Field(default="openai", description="LLM provider: 'ollama' or 'openai'")
+    provider_config_json: str = Field(
+        default="",
+        description="JSON object containing provider, cache, and observability settings for production.",
+    )
     openai_api_key: str = Field(default="", description="OpenAI API key")
     openai_model: str = Field(default="gpt-4o-mini", description="OpenAI model name")
     openrouter_api_key: str = Field(default="", description="OpenRouter API key")
@@ -331,6 +359,13 @@ def parse_cors_origins(cls, v: object) -> object:
     )
 
     # Billing / Stripe
+    billing_config_json: str = Field(
+        default="",
+        description=(
+            "JSON object containing Stripe secret_key, webhook_secret, and pro_price_id. "
+            "Used in production to keep billing credentials in one secret version."
+        ),
+    )
     stripe_secret_key: str = Field(default="", description="Stripe secret key.")
     stripe_webhook_secret: str = Field(default="", description="Stripe webhook signing secret.")
     stripe_pro_price_id: str = Field(default="", description="Stripe price id for the pro plan.")
@@ -347,6 +382,79 @@ def parse_cors_origins(cls, v: object) -> object:
         description="Return URL from Stripe customer portal.",
     )
 
+    @model_validator(mode="after")
+    def apply_provider_config(self) -> "Settings":
+        """Load bundled provider/cache credentials while preserving local split-variable support."""
+        if not self.provider_config_json.strip():
+            return self
+        values = _parse_bundled_secret(
+            self.provider_config_json,
+            "PROVIDER_CONFIG_JSON",
+            ("openai_api_key", "tavily_api_key"),
+        )
+        for name, value in values.items():
+            setattr(self, name, value)
+        redis_url = json.loads(self.provider_config_json).get("redis_url")
+        if redis_url is not None:
+            if not isinstance(redis_url, str) or not redis_url:
+                raise ValueError("PROVIDER_CONFIG_JSON redis_url must be a non-empty string.")
+            self.redis_url = redis_url
+        langfuse = json.loads(self.provider_config_json)
+        langfuse_fields = {
+            "langfuse_public_key": "langfuse_public_key",
+            "langfuse_secret_key": "langfuse_secret_key",  # nosec B105 — schema field name, not a credential
+            "langfuse_base_url": "langfuse_host",
+        }
+        for source, target in langfuse_fields.items():
+            value = langfuse.get(source)
+            if value is not None:
+                if not isinstance(value, str) or not value:
+                    raise ValueError(f"PROVIDER_CONFIG_JSON {source} must be a non-empty string.")
+                setattr(self, target, value)
+        langsmith_fields = {
+            "langsmith_api_key": "langsmith_api_key",
+            "langsmith_project": "langsmith_project",
+            "langsmith_endpoint": "langsmith_endpoint",
+            "langsmith_redaction_mode": "langsmith_redaction_mode",
+        }
+        for source, target in langsmith_fields.items():
+            value = langfuse.get(source)
+            if value is not None:
+                if not isinstance(value, str) or not value:
+                    raise ValueError(f"PROVIDER_CONFIG_JSON {source} must be a non-empty string.")
+                setattr(self, target, value)
+        langsmith_tracing = langfuse.get("langsmith_tracing")
+        if langsmith_tracing is not None:
+            if isinstance(langsmith_tracing, bool):
+                self.langsmith_tracing = langsmith_tracing
+            elif isinstance(langsmith_tracing, str) and langsmith_tracing.lower() in {"true", "false"}:
+                self.langsmith_tracing = langsmith_tracing.lower() == "true"
+            else:
+                raise ValueError("PROVIDER_CONFIG_JSON langsmith_tracing must be true or false.")
+        langsmith_sampling_rate = langfuse.get("langsmith_sampling_rate")
+        if langsmith_sampling_rate is not None:
+            try:
+                self.langsmith_sampling_rate = float(langsmith_sampling_rate)
+            except (TypeError, ValueError) as exc:
+                raise ValueError(
+                    "PROVIDER_CONFIG_JSON langsmith_sampling_rate must be a number."
+                ) from exc
+        return self
+
+    @model_validator(mode="after")
+    def apply_billing_config(self) -> "Settings":
+        """Load bundled Stripe credentials while preserving local split-variable support."""
+        if not self.billing_config_json.strip():
+            return self
+        values = _parse_bundled_secret(
+            self.billing_config_json,
+            "BILLING_CONFIG_JSON",
+            ("stripe_secret_key", "stripe_webhook_secret", "stripe_pro_price_id"),
+        )
+        for name, value in values.items():
+            setattr(self, name, value)
+        return self
+
     # Redis cache
     redis_url: str = Field(
         default="",

src/inngest_client.py

@@ -2,12 +2,34 @@
 
 from __future__ import annotations
 
+import json
 import os
 from collections.abc import Mapping
 
 import inngest
 import inngest.fast_api
 
+
+def _apply_bundled_inngest_config() -> None:
+    """Expose bundled credentials under the names expected by the Inngest SDK."""
+    raw = os.environ.get("INNGEST_CONFIG_JSON", "").strip()
+    if not raw:
+        return
+    try:
+        config = json.loads(raw)
+    except json.JSONDecodeError as exc:
+        raise ValueError("INNGEST_CONFIG_JSON must be valid JSON.") from exc
+    if not isinstance(config, dict):
+        raise ValueError("INNGEST_CONFIG_JSON must be a JSON object.")
+    for name in ("inngest_event_key", "inngest_signing_key"):
+        value = config.get(name)
+        if not isinstance(value, str) or not value:
+            raise ValueError(f"INNGEST_CONFIG_JSON must include a non-empty {name} value.")
+        os.environ[name.upper()] = value
+
+
+_apply_bundled_inngest_config()
+
 _is_production = os.environ.get("INNGEST_DEV", "").strip() != "1"
 
 inngest_client = inngest.Inngest(