Skip to content

[Automated] Draft docs (kagent): feat: Send PDF and Office files to Responses API models end-to-end - #648

Open
github-actions[bot] wants to merge 1 commit into
mainfrom
pr-tracker-draft-kagent-3200
Open

github-actions[bot] wants to merge 1 commit into
mainfrom
pr-tracker-draft-kagent-3200

Conversation

@github-actions

@github-actions github-actions Bot commented Oct 26, 2026 •

Copy link
Copy Markdown
Contributor

kagent-dev/kagent#3200 — feat: Send PDF and Office files to Responses API models end-to-end

Docs versions: 1.x (first shipped in v1.0.0-alpha16)

No code-path rule covers 3 touched file(s). The planner could route these only by identifier search, so the candidate pages may miss what they changed, including in the other mode:

  • go/adk/pkg/fileextract/llm.go
  • ui/package.json
  • ui/yarn.lock

What changes for users: In the UI and A2A clients, users can send PDF, Word, Excel, and PowerPoint files to agents that use an OpenAI or Azure OpenAI ModelConfig with apiFormat: responses. Public file links can reach Responses API and Gemini models as URL parts, while other models keep the URL as text for a fetch tool.

What the docs now say: The OpenAI provider page now explains file and link handling for openAI.apiFormat: responses, including the 50 MB request cap, the supported inline file types, the unsupported-link error, and how named attachments are represented. The A2A guide now states that the file part shape also applies to SendStreamingMessage.

Not verified: Configuration is unrun in this job; internal links are unchecked.

How kagent-3200 was drafted, and what was not verified

Plan, and what changed it

  • Planned: Document PDF, Office, and file-link behavior on the UI, A2A, and model-provider pages named by the work-list.
  • Learned: launch-ui.md and a2a-agents.md already covered most file behavior, but openai.md did not describe openAI.apiFormat: responses file input and a2a-agents.md did not name SendStreamingMessage.
  • Did: Added an OpenAI provider file-input section and a streaming clarification in the A2A guide.

Why a documentation change is necessary

The pull request changes how file inputs reach models. The UI now accepts PDF and Office attachments, links in message text can become A2A URL parts, and the OpenAI Responses API path sends PDF and Office bytes as provider file input instead of falling back to extracted text. The existing UI and A2A pages covered much of the reader workflow, but the OpenAI provider page did not say what apiFormat: responses changes for files.

What changed on disk

  • docs-site/content/kagent/1.x/examples/a2a-agents.md: Clarifies that file parts use the same shape with SendStreamingMessage.
  • docs-site/content/kagent/1.x/setup/model-providers/openai.md: Adds OpenAI Responses API file-input behavior, supported file types, link handling, and the 50 MB request cap.

How to verify this change

  1. Create an OpenAI ModelConfig with openAI.apiFormat: responses, and create an Agent that references it.
  2. In the UI or through an A2A SendStreamingMessage request, send a PDF or .docx, .xlsx, or .pptx file. The model should answer from the file contents instead of receiving the could-not-be-read note.
  3. Send a public http:// or https:// file link in the message text. A Responses API or Gemini model should fetch the link on that turn; a model that cannot fetch links should receive the URL as text.

Proposed release note

No proposed-release-note.md: markers.notes_placeable is false.

PDF and Office file input for Responses API models

The kagent UI and A2A clients can now send PDF, Word, Excel, and PowerPoint files to agents that use OpenAI or Azure OpenAI with apiFormat: responses. Supported public file links are sent as URL parts for Responses API and Gemini models, while other models keep the URL as text for a fetch tool.

What was not verified

  • Internal links are unchecked because Hugo was not run.
How the pages were chosen (triage report)

Feature clusters

These clusters look incomplete. Each names a pull request this run never saw, so documenting the cluster as it stands may describe half a feature — the classic failure is a guide telling a reader to set a field that a later change gave a default. Read the absent pull request before drafting:

Cluster Missing What it is Why it matters
c1 kagent-dev/kagent#357 show the memory retrieval indicator in the ui (merged) also worked issue #350, which kagent-dev/kagent#3200 closes

Draft a documentation change

kagent-dev/kagent#3200 — feat: Send PDF and Office files to Responses API models end-to-end

  • Product: kagent (upstream) → kagent-dev/website
  • Docs version: 1.x
  • Summary (read from the body): 🤖 written by Claude (start)
  • Not tested: no cluster was available, so every command in this diff is unrun
  • No rule covers 3 touched file(s). The candidate pages below come from the files that DID match a rule, so they may scope this narrower than the change is. Read these before deciding the change is only about the pages listed:
    • go/adk/pkg/fileextract/llm.go
    • ui/package.json
    • ui/yarn.lock

Draft branch: pr-tracker-draft-kagent-3200

…esponses API models end-to-end

Signed-off-by: GitHub Action <action@github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant