feat: gate uploads when ingestion application is disabled

CHANGELOG.md

@@ -2,6 +2,17 @@
 
 ## [Unreleased]
 
+### Added
+
+- **Optional-ingestion upload capability.** Honor
+  `DEPLOY_INGESTION_CONTAINER_APP` independently of the chat backend: `false`
+  hides standalone and Entra Copilot upload controls and rejects attachments
+  before ingestion URL or Dapr resolution. Missing preserves legacy upload
+  eligibility; `true` retains existing authentication and conversation checks;
+  malformed boolean values fail with a configuration error. Anonymous Copilot
+  uploads remain disabled. Text chat and retrieval are unchanged; no native
+  Foundry IQ conversation upload path is introduced.
+
 ## [v3.2.2] - 2026-10-14
 
 ### Fixed

README.md

@@ -165,6 +165,16 @@ rejects first-turn uploads rather than fabricating an ID. A resumed Chainlit
 thread restores context from its ordered messages and obtains a fresh managed
 ID from the hosted runtime.
 
+Uploads also require the independent `DEPLOY_INGESTION_CONTAINER_APP` capability
+from the existing App Configuration provider. `false` hides upload controls in
+standalone and Entra Copilot modes and rejects attachments before ingestion URL
+or Dapr resolution with “Attachments are unavailable in this deployment.”
+Missing preserves legacy eligibility; `true` retains existing authentication and
+conversation checks. Invalid boolean values fail with a configuration error.
+Text chat and retrieval are unchanged; this does not add native Foundry IQ
+conversation uploads. Supported ingestion sources and deployment combinations
+remain owned by the [platform documentation](https://azure.github.io/AI-Landing-Zones/agent-landing-zone/).
+
 Live acceptance still requires Basic and isolated topology validation,
 two-turn managed conversation continuity, and a two-user negative
 document-authorization test. These deployment and identity checks cannot be

docs/copilot-embedding.md

@@ -13,6 +13,16 @@ Embedding is disabled by default. Enabling it without a valid
 mode from missing Entra configuration and never downgrades an Entra failure to
 anonymous access.
 
+Entra Copilot upload controls also require the independent
+`DEPLOY_INGESTION_CONTAINER_APP` capability read through the existing App
+Configuration provider. A missing setting preserves legacy upload eligibility;
+`true` keeps the existing Entra and conversation checks. `false` disables the
+controls and server-side attachment processing, including ingestion URL/Dapr
+fallbacks, with “Attachments are unavailable in this deployment.” Malformed
+boolean values fail with an actionable configuration error. Anonymous Copilot
+uploads remain disabled regardless of this flag. Text chat, authentication and
+retrieval are unchanged; no native conversation upload path is introduced.
+
 ## Rendering and token flow
 
 Chainlit 2.12.0 Copilot mode renders a floating widget in an open Shadow DOM. It

src/gpt_rag_ui/auth/embed_security.py

@@ -21,6 +21,8 @@
     reset_request_copilot_session,
 )
 from gpt_rag_ui.config.embed_config import EmbedSettings
+from gpt_rag_ui.config.appconfig import load_ingestion_uploads_enabled
+from gpt_rag_ui.config.dependencies import get_config
 from gpt_rag_ui.telemetry.boundary_diagnostics import log_boundary_failure
 
 
@@ -654,6 +656,7 @@ def _inject_chainlit_cookie(
 def _copilot_project_settings_send(send, auth_mode: str):
     """Expose only the capabilities supported by this Copilot auth mode."""
 
+    ingestion_uploads_enabled = load_ingestion_uploads_enabled(get_config())
     pending_start = None
 
     async def send_response(message):
@@ -678,7 +681,7 @@ async def send_response(message):
                         .get("spontaneous_file_upload", {})
                     )
                     if isinstance(upload, dict):
-                        upload["enabled"] = auth_mode == "entra"
+                        upload["enabled"] = ingestion_uploads_enabled and auth_mode == "entra"
                     body = json.dumps(
                         payload,
                         separators=(",", ":"),

src/gpt_rag_ui/bootstrap.py

@@ -12,7 +12,7 @@
 from fastapi import FastAPI, Response
 from fastapi.responses import JSONResponse, StreamingResponse
 
-from gpt_rag_ui.config.appconfig import AppConfigClient
+from gpt_rag_ui.config.appconfig import AppConfigClient, load_ingestion_uploads_enabled
 from gpt_rag_ui.telemetry.boundary_diagnostics import log_boundary_failure
 from gpt_rag_ui.clients.blob import BlobClient
 from gpt_rag_ui.config.dependencies import get_config
@@ -154,21 +154,27 @@ def _format_release_value(value: str | None, missing_message: str) -> str:
     return missing_message
 
 
-def _want_chainlit_spontaneous_file_upload(auth_state: AuthState) -> bool:
-    """True whenever the effective allow_anonymous is false.
+def _want_chainlit_spontaneous_file_upload(
+    auth_state: AuthState, *, ingestion_uploads_enabled: bool = True,
+) -> bool:
+    """Enable uploads only with ingestion capability and authenticated callers.
 
     Per-conversation uploads require an authenticated caller (each file is bound to the
-    user's conversation), so we enable the Chainlit paperclip whenever auth is in force,
+    user's conversation), so the Chainlit paperclip also requires auth in force,
     regardless of whether ALLOW_ANONYMOUS came from env, App Config, or the default.
     """
 
-    return not auth_state.allow_anonymous
+    return ingestion_uploads_enabled and not auth_state.allow_anonymous
 
 
-def _sync_chainlit_spontaneous_file_upload(auth_state: AuthState) -> None:
+def _sync_chainlit_spontaneous_file_upload(
+    auth_state: AuthState, *, ingestion_uploads_enabled: bool = True,
+) -> None:
     """Write [features.spontaneous_file_upload].enabled before `import chainlit` (Chainlit reads TOML on import)."""
 
-    want_enabled = _want_chainlit_spontaneous_file_upload(auth_state)
+    want_enabled = _want_chainlit_spontaneous_file_upload(
+        auth_state, ingestion_uploads_enabled=ingestion_uploads_enabled,
+    )
     path = get_asset_root() / ".chainlit" / "config.toml"
     if not path.is_file():
         logger.warning("Chainlit config not found at %s; skipping spontaneous file upload sync", path)
@@ -728,10 +734,13 @@ def _create_chainlit_app(
     """
 
     embed_settings = embed_settings or EmbedSettings()
+    ingestion_uploads_enabled = load_ingestion_uploads_enabled(config)
     _configure_embed_environment(embed_settings)
     _configure_auth_environment(config, auth_state)
     effective_auth = auth_state or _evaluate_auth_state(config, embed_settings)
-    _sync_chainlit_spontaneous_file_upload(effective_auth)
+    _sync_chainlit_spontaneous_file_upload(
+        effective_auth, ingestion_uploads_enabled=ingestion_uploads_enabled,
+    )
 
     download_tokens = None
     if embed_settings.enabled:
@@ -757,6 +766,14 @@ def _create_chainlit_app(
     # origins can be applied in memory without modifying the deployment files.
     from chainlit.config import config as chainlit_config
 
+    # Also enforce the capability if TOML could not be written or Chainlit
+    # was imported earlier in this process.
+    if chainlit_config.features.spontaneous_file_upload is not None:
+        chainlit_config.features.spontaneous_file_upload.enabled = (
+            _want_chainlit_spontaneous_file_upload(
+                effective_auth, ingestion_uploads_enabled=ingestion_uploads_enabled,
+            )
+        )
     configure_chainlit_allowed_origins(embed_settings, chainlit_config)
     import chainlit.server as chainlit_server
 

src/gpt_rag_ui/clients/ingestion_client.py

@@ -8,6 +8,11 @@
 
 import httpx
 
+from gpt_rag_ui.config.appconfig import (
+    ATTACHMENTS_UNAVAILABLE_MESSAGE,
+    load_ingestion_uploads_enabled,
+)
+from gpt_rag_ui.config.dependencies import get_config
 from gpt_rag_ui.clients.orchestrator_client import (
     _format_outgoing_request_debug,
     _get_config_value,
@@ -189,6 +194,9 @@ async def _build_ingest_documents_payload(conversation_id: str, question_id: str
 
 
 async def ingest_files_session(conversation_id: str, question_id: str, auth_info: dict, files: list[dict]) -> bool:
+    if not load_ingestion_uploads_enabled(get_config()):
+        raise RuntimeError(ATTACHMENTS_UNAVAILABLE_MESSAGE)
+
     url, target_context = _build_ingestion_url()
     headers = {"Content-Type": "application/json"}
 

src/gpt_rag_ui/config/appconfig.py

@@ -1,7 +1,7 @@
 import os
 import logging
 
-from typing import Dict, Any
+from typing import Dict, Any, Protocol
 from azure.identity import ChainedTokenCredential, ManagedIdentityCredential, AzureCliCredential
 from azure.identity.aio import ChainedTokenCredential as AsyncChainedTokenCredential, ManagedIdentityCredential as AsyncManagedIdentityCredential, AzureCliCredential as AsyncAzureCliCredential
 from azure.appconfiguration import AzureAppConfigurationClient
@@ -20,6 +20,29 @@
 APP_LABEL = 'agent-app-ui'
 LEGACY_APP_LABEL = 'gpt-rag-ui'
 
+ATTACHMENTS_UNAVAILABLE_MESSAGE = "Attachments are unavailable in this deployment."
+
+
+class IngestionUploadConfig(Protocol):
+    def get(self, key: str, default: str, type: type) -> str: ...
+
+
+def load_ingestion_uploads_enabled(config: IngestionUploadConfig) -> bool:
+    """Read the independent ingestion capability without changing provider precedence.
+
+    Missing preserves legacy upload eligibility; a present malformed value is
+    an operator error, never an implicit opt-in.
+    """
+    value = config.get("DEPLOY_INGESTION_CONTAINER_APP", "true", str).strip().lower()
+    if value in {"true", "1", "yes", "y", "on"}:
+        return True
+    if value in {"false", "0", "no", "n", "off"}:
+        return False
+    raise ConfigurationError(
+        "DEPLOY_INGESTION_CONTAINER_APP must be a boolean (true or false). "
+        "Set false for deployments without the ingestion service."
+    )
+
 
 def candidate_keys(key: str) -> list[str]:
     """Return the lookup keys for ``key``; only the exact key name is read."""

src/gpt_rag_ui/services/chat.py

@@ -20,6 +20,10 @@
     replace_source_reference_links,
 )
 from gpt_rag_ui.config.dependencies import get_config
+from gpt_rag_ui.config.appconfig import (
+    ATTACHMENTS_UNAVAILABLE_MESSAGE,
+    load_ingestion_uploads_enabled,
+)
 from gpt_rag_ui.clients.blob import BlobClient
 from gpt_rag_ui.services.conversation_security import get_owned_conversation
 from gpt_rag_ui.clients.ingestion_client import ingest_files_session
@@ -526,12 +530,18 @@ def _trim_for_log(value: str, limit: int = 400) -> str:
         rejected: list[str] = []
         total_declared_bytes = 0
         file_reply_parts: list[str] = []
+        ingestion_uploads_enabled = True
 
         if message.elements:
+            ingestion_uploads_enabled = load_ingestion_uploads_enabled(config)
             for element in message.elements:
                 if not isinstance(element, cl.File):
                     continue
 
+                if not ingestion_uploads_enabled:
+                    rejected.append(ATTACHMENTS_UNAVAILABLE_MESSAGE)
+                    continue
+
                 mime = (getattr(element, "mime", "") or "").lower()
                 name = getattr(element, "name", "upload")
                 path = getattr(element, "path", None)
@@ -661,7 +671,10 @@ def _trim_for_log(value: str, limit: int = 400) -> str:
                 return
 
         user_ask = (message.content or "").strip()
-        if CHAT_BACKEND == "hosted_agent" and not user_ask:
+        if not user_ask and (
+            CHAT_BACKEND == "hosted_agent"
+            or (rejected and not ingestion_uploads_enabled)
+        ):
             final_text = "".join(file_reply_parts).strip()
             if not final_text:
                 final_text = "Enter a question to start or continue the hosted conversation."
@@ -670,7 +683,7 @@ def _trim_for_log(value: str, limit: int = 400) -> str:
             response_msg.content = final_text
             await response_msg.update()
             logger.info(
-                "Skipping hosted agent (empty ask): conversation=%s question_id=%s",
+                "Skipping chat backend (empty ask): conversation=%s question_id=%s",
                 hosted_conversation_id or "new",
                 message.id,
             )