fix: load complete paginated run history

[Eilen6316]

Summary

Fixes #3290.

This PR fixes incomplete chat history restoration for long runs where a single run has more message events than the first /messages page returns. Previously the frontend loaded only one page per run, so older messages could disappear after reopening/restarting the app, and exports based on the loaded message list could miss the same older context even though token usage still reflected the full run.

Changes

• Continue loading /api/threads/{thread_id}/runs/{run_id}/messages pages for the same run while the backend returns has_more.
• Use the oldest returned message seq as the next before_seq cursor, so long runs are restored page by page instead of being marked loaded after the first page.
• Add seq to the frontend RunMessage type and helper functions for pagination cursor handling / URL construction.
• Keep compatibility with both has_more and the older camelCase hasMore response field.
• Add a shared backend trim_run_message_page() helper so both run-message endpoints trim the extra limit + 1 sentinel row consistently.
• Preserve pagination boundaries correctly:
  • default/latest and before_seq pages keep the newest side and drop the older sentinel row;
  • after_seq pages keep the oldest side and drop the newer sentinel row.
• Add regression tests for backend pagination boundaries and frontend cursor helpers.

Risk / scope

The change is limited to run-message history pagination. It does not alter run creation, streaming, model invocation, or message persistence. The backend behavior keeps the existing limit + 1 contract and only centralizes which sentinel row is trimmed.

Tests

• Backend pagination tests:

  docker exec deer-flow-gateway sh -lc 'cd /app/backend && uv run pytest tests/test_thread_run_messages_pagination.py tests/test_runs_api_endpoints.py'

  Result: 27 passed, 1 warning

• Frontend unit tests:

  docker exec deer-flow-frontend sh -lc 'cd /app/frontend && pnpm test -- --run tests/unit/core/threads/message-merge.test.ts'

  Result: 20 passed, 128 passed

• Backend lint / format:

  docker exec deer-flow-gateway sh -lc 'cd /app/backend && uvx ruff check app/gateway/routers/runs.py app/gateway/routers/thread_runs.py app/gateway/pagination.py tests/test_runs_api_endpoints.py tests/test_thread_run_messages_pagination.py tests/_run_message_pagination_helpers.py && uvx ruff format --check app/gateway/routers/runs.py app/gateway/routers/thread_runs.py app/gateway/pagination.py tests/test_runs_api_endpoints.py tests/test_thread_run_messages_pagination.py tests/_run_message_pagination_helpers.py'

  Result: passed

• Frontend format:

  docker exec deer-flow-frontend sh -lc 'cd /app/frontend && pnpm exec prettier --check src/core/threads/hooks.ts src/core/threads/types.ts tests/unit/core/threads/message-merge.test.ts'

  Result: passed

[WillemJiang]

@Eilen6316 Do you mind resolving the confilct?

[Copilot]

Pull request overview

This PR fixes incomplete restoration of long run histories by paginating through /api/threads/{thread_id}/runs/{run_id}/messages until the backend indicates there are no more pages. It also centralizes backend “limit + 1” sentinel trimming so pagination boundaries remain correct for both forward (after_seq) and backward/latest (before_seq / default) pagination.

Changes:

• Frontend: add run-message pagination cursor support (seq), helpers to detect has_more/hasMore, compute the next before_seq, and build message URLs safely.
• Frontend: update useThreadHistory to repeatedly fetch additional pages for the same run using before_seq until fully loaded.
• Backend: introduce a shared trim_run_message_page() helper to consistently trim the sentinel row while preserving correct page boundaries; add regression tests for trimming behavior.

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
frontend/tests/unit/core/threads/message-merge.test.ts	Adds unit tests for new run-message pagination helpers and URL construction.
frontend/src/core/threads/types.ts	Extends RunMessage to include optional seq used for pagination cursors.
frontend/src/core/threads/hooks.ts	Implements cursor-based multi-page loading per run and adds pagination helper utilities.
backend/tests/test_thread_run_messages_pagination.py	Adds boundary regression tests for thread-scoped run message pagination trimming.
backend/tests/test_runs_api_endpoints.py	Adds boundary regression tests for run-scoped message pagination trimming.
backend/tests/_run_message_pagination_helpers.py	Introduces a shared assertion helper for run-message pagination tests.
backend/app/gateway/routers/thread_runs.py	Switches to shared trimming helper for run-message pagination.
backend/app/gateway/routers/runs.py	Switches to shared trimming helper for run-message pagination.
backend/app/gateway/pagination.py	Adds trim_run_message_page() to correctly trim “limit + 1” pages for both cursor directions.

[Eilen6316]

@Eilen6316 Do you mind resolving the confilct?

Sure, I’ve resolved the conflict. Could you please review it again when you get a chance? Thanks!