Generative AI topics

Before dropping a column, check whether Genie has been using it

ivanvyd — Wed, 16 Sep 2026 19:14:29 GMT

I was reviewing a schema change recently and wanted a quick answer to a simple question:

Could this column be used by a Genie Agent somewhere without me realizing it?

There's now a pretty useful way to check I want to share with you.

Databricks' March 25, 2026 release notes introduced entity_metadata.genie_space_id in both system.access.column_lineage and system.access.table_lineage. The field identifies the Genie Agent when a lineage record originates from it querying data (see March 2026 | Databricks on AWS).

For a column such as customer_segment, I can start with this query:

SELECT
    workspace_id,
    entity_metadata.genie_space_id AS genie_space_id,
    direct_access,
    COUNT(DISTINCT event_id) AS observed_events,
    MAX(event_time) AS last_seen
FROM system.access.column_lineage
WHERE source_table_full_name = 'prod.sales.orders'
  AND source_column_name = 'customer_segment'
  AND entity_metadata.genie_space_id IS NOT NULL
  AND event_date >= date_sub(current_date(), 90)
GROUP BY
    workspace_id,
    entity_metadata.genie_space_id,
    direct_access
ORDER BY last_seen DESC;

I can run this from a Unity Catalog-enabled workspace where I have access to the lineage system tables.

I use the result set to inspect observed usage by workspace, Genie ID, and direct-access status. observed_events counts distinct lineage events. I do not treat it as a count of user questions or conversations. last_seen gives me the latest recorded lineage timestamp inside the selected window.

I keep indirect access in the result

direct_access = false means the source is a dependency discovered through view expansion.

A Genie Agent can query a view while the recorded lineage points to an underlying column. If I filter the query to direct access, I miss those records.

That distinction matters during a schema review because the team changing the base table may never see the base table referenced in the Genie-facing query.

An empty result does not prove the change is safe

I treat an empty result as absence of matching recorded lineage in the window I queried. I do not treat it as proof that nothing depends on the column.

A Genie Agent may not have used the column during the lookback period. Databricks emits lineage records when it can infer lineage, so the lineage tables do not capture every read or write event.

I also account for the scope and latency of the system tables. These lineage tables cover workspaces in the account within the same cloud region, and system-table updates are not real-time. Recent activity may not appear yet.

The SQL does not perform recursive dependency traversal through downstream tables that separate jobs populate.

A chain such as:

orders > ETL job > summary table > Genie

needs a broader dependency review. This query does not connect those separate lineage events for me.

I pair observed usage with configuration review

For higher-risk changes, I also inspect Genie configuration, including relevant agents that did not appear in the lineage results.

I check example SQL, instructions, join expressions, and referenced views or functions.

The Genie API can return a serialized configuration through:

GET /api/2.0/genie/spaces/{space_id}?include_serialized_space=true

Requesting that export requires at least CAN EDIT permission on the agent. The exported configuration includes data sources and instructions, including example SQL and function references.

I do not stop at the attached-table list. Genie can query other tables for which it has Unity Catalog permissions, including tables referenced in instructions or generated queries.

For table renames or drops, I use the same approach with system.access.table_lineage: I change the system table and remove the source_column_name predicate. The Genie metadata is available there as well.

My deployment review follows this sequence:

Proposed schema change > inspect Genie lineage > review dependencies and configurations > test > deploy

I use the lineage query as the starting point for investigating observed dependencies. It is evidence for the review, not automatic approval to change the schema.

How do you test whether a retriever stops too early?

ivanvyd — Fri, 11 Sep 2026 12:12:48 GMT

Hello everyone,

Databricks’ Adaptive Instructed-Retriever announcement⁠ describes a model that learns when to stop searching. How would you test whether it stops after finding a general policy but before finding the exception that changes the answer?

I’d consider an offline test using questions with reviewed supporting evidence. Where the implementation allows it, continue early-stopped runs from the same history with a small extra search budget. Keep the model, corpus and final passage count fixed, then measure how often further search recovers missing evidence.

I’d measure evidence coverage separately from answer quality. An unsuccessful extra search wouldn’t prove the original retrieval was complete, either.

Has anyone tried something similar? What baseline or metric helped you judge whether the latency savings justified the risk of missing evidence? I am very eager to get your thoughts.

Enterprise workspace blocked by Databricks-set rate limit of 0 on all models (AWS Marketplace)

khanzada — Thu, 10 Sep 2026 05:31:14 GMT

I'm on an Enterprise plan (billed via AWS Marketplace, active payment method, $399 credit remaining) and every model in my Unity AI Gateway — glm-5-3-flash, claude-opus-5, grok-4-6, gpt-5-6-sol, etc. — returns:

"error_code":"PERMISSION_DENIED","message":"PERMISSION_DENIED: The endpoint is temporarily disabled due to a Databricks-set rate limit of 0."

The workspace. I've checked:
- AI Gateway rate limits on individual models — not configured (blank, not 0)
- Budgets tab — "No budgets apply"
- Plan tier — confirmed Enterprise, not trial/Free Edition

Setting my own rate limits in the UI doesn't override this either. This looks like the same Databricks-imposed 0-rate-limit issue described in an earlier thread (https://community.databricks.com/t5/generative-ai/permission-denied-the-endpoint-is-temporarily-disabled-due-to-a/td-p/150773), and also matches other recent reports of this happening on paid/Premium accounts, not just trials.

I don't have a support subscription to open a ticket directly. Can someone from Databricks help get my workspace moved off whatever trust tier is causing this?

Thanks,

Genie Agent response Export to PDF via API

GunaR — Tue, 08 Sep 2026 09:39:19 GMT

Hello,

Got a need to invoke the Genie Agent chat (in agent mode) via API, and on the request completion, export the agent response in PDF and share the exported PDF in email.

Am already using 'Agent Mode Create Response' endpoint to invoke the chat programmatically. https://docs.databricks.com/api/genie/v1/agent-mode-create-response

What's the best way to export the response in PDF? I know there is no explicit endpoint to achieve the same.

Thanks,
Guna.

Genie Agent content search on a managed volume fails with "The request was invalid"

mramanindia — Fri, 04 Sep 2026 14:08:06 GMT

Hi all,

I am trying to enable content search on a Unity Catalog managed volume so a Genie Agent can answer questions over PDFs. The Enable action fails every time with:

"Failed to update content search - The request was invalid."

Environment: AWS, eu-west-2 (London), serverless workspace on Default Storage, pay-as-you-go. This started as a trial and the failure is identical before and after adding a payment method.

What I have verified:

- Preview "Analyze Files in Volumes with Genie Agents" is enabled.

- The volume is a managed volume. Catalog binding is "All workspaces have access" (OPEN). Reproduced on two catalogs, including a freshly created one.

- I have CAN MANAGE on the volume.

- ai_parse_document() runs successfully on the PDFs in the volume.

- Foundation Model APIs work (ai_query succeeds).

- Lakebase is available in the region.

- I can create a Vector Search endpoint via the API and it comes ONLINE.

- Files are 4 PDFs, each under 10 MB.

Without content search the agent does not retrieve the files ("no relevant content from the reports directory"), so document and hybrid questions are not possible.

Questions:

1. Is content search supported on Default Storage / serverless-only workspaces? It is not listed under the Default Storage limitations, but it is the one variable I cannot rule out.

2. Is there a way to get a more detailed error than "The request was invalid"? No public API seems to expose this.

Happy to share the request ID from the browser network log if that helps.

Thanks!

Databricks-Native AI Agent for Job Incident Detection, RCA & Safe Remediation

VibinRoy_C — Sat, 05 Sep 2026 11:29:57 GMT

I’m exploring an architecture for a Databricks-native AI Agent for intelligent Spark job incident detection, root cause analysis (RCA), and safe remediation, and I would love your technical feedback.

The Problem

A common operational challenge is the "Silent Long-Runner" or stalled workload—for example, a Spark job that normally finishes in ~1 hour suddenly runs for 8+ hours with zero meaningful progress, wasting DBUs and blocking pipelines.

Currently, investigating this requires engineers to manually correlate disparate data sources:

Spark UI: Job/Stage DAGs, Task skew, stragglers, shuffle memory/disk spill.
Compute State: Driver/Executor availability, OOMs, memory pressure, thread dumps.
Logs & I/O: Driver/executor logs and real-time input/output progression.
Historical Context: Baseline metrics from past successful runs.

Proposed Concept: Execution-Aware Agentic Intelligence

Instead of a simple log-parsing chatbot, the goal is a closed-loop agent framework built on Databricks-native capabilities (System Tables, MLflow, Agent Framework/MCP, Unity Catalog):

Baseline Learning: Dynamically detects anomalies against historical execution patterns instead of static time limits.
Multi-Signal RCA & Evidence Chain: Correlates live Spark UI metrics, logs, and compute telemetry to yield an explainable diagnosis (e.g., "Root Cause: Executor OOM & Disk Spill Skew — 92% Confidence").
Policy-Governed Remediation: Uses an Observe → Recommend → Auto-Remediate model, requiring Human-in-the-Loop (HITL) approval for high-risk actions (e.g., job cancellation, cluster resizing).
Closed Feedback Loop: Captures whether actions resolved the issue to continuously improve accuracy and quantify DBU/cost savings.

Questions for the Community

Telemetry Access: How much granular Spark UI data (stage progress, task skew, shuffle details) can be programmatically accessed in real time via System Tables or APIs?
Native Pattern: What native capabilities would you recommend for correlating these runtime and log signals efficiently?
Remediation Safety: What guardrails or design patterns do you recommend for automated or HITL actions on running jobs?
Prior Art: Has anyone built or tested a similar agentic incident detection workflow inside Databricks?

Would love to hear your thoughts, feedback, or architectural suggestions!

#genAI #dataengineering #AIagent

Genie - Cost monitoring and usage

Sanjeeb2024 — Sat, 05 Sep 2026 05:42:31 GMT

Hi Team,

Genie is widely used within our department, and we would like to establish appropriate governance around its usage. Do we have any existing guidelines or recommended practices for this? In particular, we are looking for guidance on:

Cost monitoring: How can we monitor Genie usage, including who is using it and the associated consumption or cost?
Usage limits: Is it possible to set an upper limit or other controls to restrict Genie usage?Any relevant documentation, recommendations, or examples would be appreciated.

Regards - Sanjeeb

Can Unity AI Gateway Model Service guardrails protect an existing LangGraph agent deployed as a Data

HariUmeshNaraya — Thu, 03 Sep 2026 06:51:04 GMT

I have a LangGraph agent deployed as a Unity Catalog model on a
Databricks Serving Endpoint.

Architecture:

User
↓
Databricks Serving Endpoint
↓
UC Model: ai_workspace.agent.agent
↓
LangGraph agent
↓
LLM + tools

The serving endpoint shows "AI Gateway enabled", but the endpoint
configuration currently only shows:

ai_gateway:
usage_tracking_config:
enabled: true

I am trying to enable the Unity AI Gateway guardrails.

In the Unity AI Gateway UI, I can create a "Model Service" and configure
guardrails. However, when selecting the destination model, my existing
agent/UC model:

ai_workspace.agent.agent does not appear.

The available destinations appear to be Databricks-hosted foundation
models such as Claude, GPT, Qwen, etc.

1. Can a Unity AI Gateway Model Service be placed in front of an existing
LangGraph agent deployed on a Databricks Serving Endpoint?

2. If not, can the Model Service be used by the LangGraph agent as the
LLM endpoint, so that guardrails can be enabled for the LLM calls?

3. Does the Unity AI Gateway j guardrail protect the entire
agent execution path, including LangGraph tool calls, or only the
request/response to the underlying model?

4. What is the recommended Databricks architecture for applying
prompt-injection/jailbreak(Guardraiils) protection to a custom LangGraph agent?

5. For an agent serving endpoint, is "AI Gateway enabled" currently
limited to usage tracking/inference tables rather than the newer
Unity AI Gateway service-policy guardrails?

I would especially appreciate clarification from someone familiar with
the current Unity AI Gateway + Agent Serving + Guardrails architecture.

How to get started with Genie Ontlogy?

Niyojit — Wed, 29 Jul 2026 07:04:40 GMT

Hello Databricks Community,

I recently started exploring Genie Ontology and would like to implement it as part of a POC. However, I'm not sure how to get started.

I have a few questions:

What are the prerequisites for using Genie Ontology?
How do I build and configure an ontology from scratch?
Are there any official guides, best practices, or sample implementations that you would recommend?
Is Genie Ontology currently in Public Preview, Private Preview, or General Availability?
Where should I be able to access it in the Databricks workspace? I don't see the Genie Ontology option or icon in my workspace, so I'm wondering if there are any workspace requirements, feature flags, or permissions that need to be enabled.

If anyone has already worked with Genie Ontology, I'd really appreciate any guidance, documentation, or tips on getting started.

Thank you!

Best Practices for Deploying Custom LLM Pipelines on Databricks?

exrwebflowai — Tue, 01 Sep 2026 10:44:42 GMT

Hi community,

We are currently designing an enterprise-grade LLM pipeline on Databricks to handle end-to-end data processing, model fine-tuning, and inference serving.

While Spark handles our data orchestration effectively, we are looking at optimizing model latency and integration with custom frontend/backend architectures. For reference, we've been following design patterns similar to those used in LLM Development Services to structure custom AI solutions.

A few questions for the group:

Serving Options: How are you balancing Databricks Model Serving with external API endpoints in high-throughput applications?
Vector Search & Storage: Are you predominantly using Databricks Vector Search, or integrating third-party vector databases for custom retriever pipelines?
Monitoring: What tools or logging setups have worked best for tracking drift and hallucinations in production?

Would love to hear how other teams are structuring their LLM architecture on the platform!

Premium pay-as-you-go account still blocked by Databricks-set rate limit of 0

LeMi-1 — Mon, 31 Aug 2026 10:08:13 GMT

Hello,

I’m having an issue using the Databricks Foundation Model API.

My account is on the Premium plan, I have a valid payment method attached, and I currently have trial credits available. However, when I try to use the GLM Foundation Model endpoint, every request returns:

403 PERMISSION_DENIED: The endpoint is temporarily disabled due to a Databricks-set rate limit of 0.

I understand from previous Databricks Community answers that this can happen when a workspace remains in the TRIAL_VERIFIED trust tier, and that adding a payment method does not necessarily move the workspace to PAYABLE_VERIFIED.

I already contacted Databricks Support (ticket #01003452), but they closed the ticket because I do not have a separate support package and recommended that I ask here instead.

This is a personal Premium pay-as-you-go account, not a company/enterprise account.

Could a Databricks employee or community engineer please help clarify:

1. Is my workspace likely still in the TRIAL_VERIFIED trust tier?
2. How can an individual Premium pay-as-you-go customer get moved to PAYABLE_VERIFIED?
3. Who can remove the Databricks-set rate limit of 0 if standard Support cannot help?
4. Is there anything I can do from the Account Console to enable the GLM Foundation Model API?

Thank you.

How Can a Databricks App Render the Same Detailed Response as the Genie UI?

GauriBhogle — Sun, 30 Aug 2026 22:35:05 GMT

I’m building a Databricks App with a Genie Space at the core and have run into a difference between the Genie UI and the Genie Conversation API.

For the same business question, the Genie web UI can produce a detailed response containing an executive summary, breakdowns, query-backed results, and visualizations.

When I invoke the same Genie Space from a Databricks App using the Databricks SDK (start_conversation_and_wait / create_message_and_wait), the returned narrative is often significantly shorter.

I also inspected the message attachments. In some Genie UI conversations, much of the detailed analysis appears to be represented through SQL/query/visualization attachments rather than a complete text attachment.

I tried retrieving query-attachment results through the Genie SDK/API, but I have not yet been able to reproduce the same detailed experience that is visible in the Genie web UI.

Questions:

Is the detailed narrative displayed in the Genie web UI always available through the Conversation API, or does the UI perform additional client-side synthesis/rendering?
What is the recommended API/SDK method for retrieving the result data associated with every query attachment returned by a Genie message?
Can Genie-generated visualizations from the web UI be retrieved/rendered in a Databricks App?
Is there a supported way for a Databricks App to request the same detailed/deep-research-style response produced by the Genie UI?
For production Genie-powered Apps, what is the recommended pattern for presenting narrative + query results + visualizations while keeping Genie as the reasoning layer?

My goal is not to execute or reconstruct Genie's business logic separately in the application. I want the App to faithfully present the answer and evidence generated by Genie.

Any guidance on the expected API behavior or recommended implementation pattern would be appreciated.

Agent mode: will per-turn context reporting be added, or is there another way to read it?

ncastagnet_mc — Fri, 28 Aug 2026 18:22:01 GMT

Genie spaces are steered by a semantic layer: text instructions (e.g. "when the user says 'sales', always filter to region = 'EMEA'") and example question→SQL pairs.

We read turn payloads programmatically — via the message endpoint behind the Genie UI (/api/2.0/data-rooms/{space}/conversations/{cid}/messages/{mid}), with the public conversation API as fallback (it's thinner and omits exactly the context below).

Comparing the two modes:

Standard mode tells you what each turn used. Every message payload records the instructions applied (progress_report.context_info.primitive_instructions), the example SQL pairs it drew on (examples.example_questions, with confidence scores), and the tables retrieval selected (relevant_tables). When an answer looks wrong, you can check whether the instruction fired before touching any SQL.
Agent mode reports none of this. The payload carries only the research trail — sub-questions, research thoughts, queries, the final report. Nothing records which instructions or example SQL shaped the turn. And this is the mode where it matters most: one deep-research turn can run 5+ SQL queries, so a missed instruction silently propagates across all of them.

Specifically, what we need: a new output item type in the agent mode API's response stream (POST /api/2.0/genie/agents/{agent_id}/responses) carrying the per-turn context standard mode already reports: primitive_instructions (instructions applied), example_questions (example SQL used, with confidence), relevant_tables (tables selected).

The space API doesn't close the gap. It exposes the configured instruction/example list — what could have steered a turn, not what did — and it's current-state only: no version history, so after an edit you can't reconstruct the config a past turn ran against. (If config versioning is planned, that would help here too.)

Questions:

Will agent mode report per-turn context (instructions / example SQL used) the way standard mode does? If not, is it on the roadmap?
Is there another way to read it today — an API field, an event type in the stream, a debug option, or the UI?

Thanks!

Custom Agent traces only showing basic input and output

taniumalloy — Mon, 24 Aug 2026 19:32:37 GMT

I've been working with MLflow Agent Server deployments using the OpenAI Agents SDK hosted both locally and in Databricks Apps. Currently, tracing is instrumented with:

set_trace_processors([]) mlflow.openai.autolog()

Locally, everything works great and my traces show annotations for tool calls and include lots of rich telemetry:

However, when I migrate to Databricks apps and try to leveraged Databricks Experiments, I lose all the details. It just shows the basic input and output text:

I've tested with both LangChain and OpenAI Agents SDK apps and nothing has changed on the Databricks side.

Is there currently limited support for tracing custom agents? Or has anyone been able to get proper tracing going with custom agents?

Following the guide here: https://github.com/databricks/app-templates/blob/main/agent-openai-agents-sdk/agent_server/agent.py

Azure Databricks Genie + Copilot Studio: MCP tools discovered and enabled, but Copilot never invokes

rquezada — Mon, 24 Aug 2026 18:38:35 GMT

Hi everyone,

I'm trying to integrate an Azure Databricks Genie Agent with Microsoft Copilot Studio through the MCP integration, and I'm stuck in a situation where everything appears correctly configured, but Copilot never actually invokes the Genie tools.

What has already been validated:

A Genie Agent called "PLANI GENIE" exists in Azure Databricks.
The Genie Agent is active and has a valid Agent ID.
The Genie Agent is attached to an active SQL Warehouse.
Under Unity AI Gateway → MCPs, the MCP appears as Active.
The MCP type is "Genie Agent".
The MCP exposes the expected tools:
- query_space_<agent_id>
- poll_response_<agent_id>
Both tools are visible in Databricks.

On the Copilot Studio side:

The Azure Databricks connection was created successfully through Power Platform using OAuth authentication
Connection status is healthy/connected.
When adding the Azure Databricks Genie tool, Copilot correctly discovers the Genie Space.
The Genie Space selector appears and I can select "PLANI GENIE".
The tool configuration shows both:
- query_space_<agent_id>
- poll_response_<agent_id>
Both tools are enabled.
Authentication mode is set to User.
The agent has been saved and published after every configuration change.

Agent model:

Claude Sonnet 4.6

Instructions are intentionally simple. The agent is instructed to use Genie for questions related to CMGs, generation, demand, policies, CAPEX, IFOR, tables and Databricks data, and to never answer from its own knowledge.

Expected behavior:

If I ask something like:

"What is the average CMG for busbar CrucEnc220?"

I would expect Copilot to:

Call query_space.
Poll using poll_response.
Return the Genie response.

Actual behavior:

Instead of invoking Genie, Copilot responds with messages such as:

"I don't have access to Azure Databricks Genie."

"I only have local tools such as bash, grep, file utilities, etc."

"The MCP tools are not available in this environment."

"I cannot access PLANI GENIE."

What confuses me is that Copilot clearly discovers:

the MCP,
the Genie Agent,
the Genie Space,
query_space,
poll_response,

yet during execution it behaves as if none of them exist.

At this point I have already verified:

✅ Genie Agent exists

✅ Genie Agent active

✅ Warehouse active

✅ MCP active

✅ MCP tools exposed

✅ Azure Databricks connection healthy

✅ Copilot discovers Genie Space

✅ Copilot discovers query_space

✅ Copilot discovers poll_response

✅ query_space enabled

✅ poll_response enabled

✅ Agent published

My questions are:

Is there any additional orchestration configuration beyond enabling query_space and poll_response?
Are there known issues where Copilot discovers the MCP tools but never actually invokes query_space?

Note: I do not have the AI Orchestration setting mentioned in the Databricks documentation. The Genie Agent and MCP tools are configured correctly, but the orchestration option is not present in my Copilot Studio environment.

Any guidance or working examples would be greatly appreciated. Thanks!

Databricks-hosted Foundation Model APIs blocked by workspace rate limit of 0

APIUser4827 — Mon, 24 Aug 2026 05:15:45 GMT

Hello,

My Databricks-hosted Foundation Model endpoints appear READY, but every runtime invocation is blocked by a workspace-level quota.

Workspace details:
- Workspace ID: 7474647458172567
- Workspace URL: https://dbc-731cb41b-741e.cloud.databricks.com
- Region: us-west-2
- Serverless compute: enabled

Reproduction:
1. GET /api/2.0/serving-endpoints/databricks-claude-sonnet-5 returns HTTP 200.
2. Endpoint state is READY and config_update is NOT_UPDATING.
3. POST /serving-endpoints/databricks-claude-sonnet-5/invocations returns HTTP 403:
PERMISSION_DENIED: The endpoint is temporarily disabled due to a Databricks-set rate limit of 0.

The same workspace-level rate-limit error has also affected other Databricks-hosted foundation models, so this does not appear to be a missing endpoint or Model Service configuration issue.

Billing has been upgraded from the trial flow and a payment method is present. Partner-powered AI features are enabled. Creating another Model Service does not resolve the problem.

Could a Databricks community engineer please clarify:
1. How can an individual pay-as-you-go workspace receive a non-zero Foundation Model API rate limit?
2. Does the workspace/account need a manual entitlement or verification-tier update?
3. Which billing or account team can remove the Databricks-set rate limit of 0 if standard support is unavailable?

Thank you.

Endpoint Error

aryan_s1977 — Fri, 21 Aug 2026 11:07:09 GMT

I have a GCP Databricks account on the Enterprise plan with GCP Marketplace configured as the default payment method. Partner-powered AI features are enabled, and we enabled the Supervisor API Beta preview. Unity Catalog v3 model services are visible under system.ai, but calls to system.ai.gpt-5-6-sol and system.ai.claude-sonnet-5 fail with PERMISSION_DENIED: The endpoint is temporarily disabled due to a Databricks-set rate limit of 0. Could you confirm why the account has a Databricks-set zero rate limit for these hosted models and whether it can be enabled?

Questions about "Store OpenTelemetry traces in Unity Catalog"

Yuki — Thu, 20 Aug 2026 23:58:33 GMT

Hi,

I'm trying to use the OpenTelemetry traces in Unity Catalog https://docs.databricks.com/aws/en/mlflow3/genai/tracing/trace-unity-catalog

But I'm having trouble to understand the behavior.

First, I can't find any data in the following two tables.

<table_prefix>_otel_logs
<table_prefix>_otel_metrics

Second, I can't understand the differences of the Archive traces to a Delta table https://docs.databricks.com/gcp/en/mlflow3/genai/eval-monitor/archive-traces .

Which one should we adopt going forward?

Thank you,

Bedrock and Genie

data_architect2 — Wed, 19 Aug 2026 17:07:55 GMT

Genie experts

We are planning to utilize the AWS Bedrock agent to work with Genie One and invoke endpoints within Databricks. Our setup will involve multiple Databricks accounts that will be integrated with the Bedrock agent. I would appreciate any guidance on best practices, as well as lessons learned from similar implementations. Thank you for your assistance

Genie Agent Volume Attachment (Beta) Docs mentions multiple formats but images aren't supported

Santhosh_23 — Tue, 18 Aug 2026 06:38:39 GMT

Hi Databricks Team,

I'm testing the Genie Agent volume attachment feature (Beta) and have a question regarding supported file formats.

According to the documentation, Genie Agents can be attached to a Volume containing various file formats. The supported formats appear to include documents and other file types that can be used as knowledge sources.

However, when I try to attach a Volume that contains image files (for example, .png or .jpg), the Databricks workspace UI indicates that images are not supported.

Could someone help clarify the following?

Are image formats (.png, .jpg, .jpeg, etc.) officially supported for Genie Agent Volume attachments in the current Beta release?
If image support is documented, is there an additional configuration or prerequisite required to enable it?
Is the workspace UI limitation expected behavior, or could this be a bug/inconsistency between the documentation and the current implementation?
If images are not yet supported, is there an updated list of the currently supported file formats for Genie Agent knowledge sources?

For reference, I am referring to the documentation that describes supported file formats for Volume attachments, but the behavior I'm seeing in the workspace seems to differ.

Any clarification would be greatly appreciated. Thanks!