feat: implement ListTasks across service, CLI, and MCP - #112
Merged
Merged
Conversation
ListTasks is one of the eleven A2A v1.0 methods and Handler implemented it on no surface; there was no way to ask an agent what tasks exist. A2AService gains list_tasks (one page, raw ListTasksResponse) and list_all_tasks (follows next_page_token to the end, with a repeated- token guard so a misbehaving server cannot loop it forever). An unset proto3 page size is indistinguishable from zero and servers reject zero, so requests always carry an explicit page size (default 50). task_state_from_label and TASK_STATE_LABELS translate compact labels like "completed" or "input-required" into TaskState values for filters. CLI: handler task list with --context-id, --status, --page-size, --history-length, and --include-artifacts. Text output is a compact one-line-per-task listing; --output json wraps the full wire-format tasks with a count; ndjson emits one task per line. All pages are always fetched rather than silently truncating at the first. MCP: a list_tasks tool with the same filters, returning count + tasks. Verified against the local streaming test agent: listing returns all tasks across pages (page-size 2 over 6 tasks), and filtering by context and by status both narrow correctly. The TUI task browser this unlocks is left as a follow-up. Closes #101 Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- list_all_tasks deduplicates tasks by ID, stops when the server returns the token that was just sent (a guaranteed loop), and caps pagination at MAX_LIST_TASKS_PAGES so a server minting fresh tokens forever cannot spin the client; the old repeated-token guard appended the replayed page's tasks before breaking, double-counting them. - page_size below 1 is rejected with a clear InputValidationError at the service, CLI, and MCP layers instead of silently becoming the default (0) or reaching the wire (negatives); the 50-task default is now stated in the CLI and MCP help. - task_state_from_label only accepts the real states in TASK_STATE_LABELS: "unspecified" maps to the proto default and would silently drop the status filter, so it errors like any unknown label. - The CLI --status option uses the tolerant parser instead of click.Choice, so the documented hyphenated spellings (input-required) and mixed case work, matching the MCP tool. - The list_tasks log no longer reports "any" for a zero status value, and the test fake raises past its scripted pages instead of replaying the last one, so extra requests fail loudly. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
alDuncanson
force-pushed
the
feat/list-tasks
branch
from
September 8, 2026 15:41
a2ed01b to
4696e7e
Compare
alDuncanson
marked this pull request as ready for review
September 8, 2026 15:43
alDuncanson
added a commit
that referenced
this pull request
Sep 8, 2026
Follow-ups from #112. list_all_tasks silently returned partial results. When a pagination defense stopped the crawl early, callers got a short list with no way to tell it apart from a complete one. It now returns a TaskListing carrying the tasks plus a truncated flag, surfaced by the CLI in all three output formats and by the MCP tool as a `truncated` field. The loop also stops on any non-progressing page rather than only on an immediately repeated token, so a server cycling two or more tokens is caught on the first page that contributes nothing instead of spinning to the page cap. Drop the hardcoded 50-task default page size. page_size and history_length are proto3 optional fields with real presence, so leaving them unset genuinely omits them and the server applies its own default; the comment claiming otherwise was wrong. Verified against a live agent. Validation moves into shared validate_page_size and validate_history_length helpers, used by the service, CLI, and MCP instead of three copies. page_size now also rejects values above 100, the spec maximum, which a server would only reject after a round trip. history_length is validated on get_task as well, which had no check. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #101
ListTasksis one of the eleven A2A v1.0 methods and Handler implemented it on no surface — there was no way to ask an agent what tasks exist.Service layer
A2AService.list_tasks(...)wraps the SDK call and returns the rawListTasksResponsefor one page, withcontext_id,status,page_size,page_token,history_length, andinclude_artifactspass-throughA2AService.list_all_tasks(...)followsnext_page_tokento the end, with a repeated-token guard so a misbehaving server cannot loop it forevertask_state_from_label/TASK_STATE_LABELStranslate compact labels likecompletedorinput-required(hyphens or underscores) intoTaskStatevaluesCLI
handler task listwith--context-id,--status(choice of real state labels),--page-size,--history-length, and--include-artifacts:id state context=...)--output json: full wire-format tasks wrapped with a count--output ndjson: one task per lineMCP
A
list_taskstool with the same filters, returning{count, tasks}in wire format.Verification
--output json task listreturned all tasks,--status input_requiredand--context-idfilters both narrowed correctly, and--page-size 2over 6 tasks still returned all 6 across 3 pagesFollow-up
The TUI task browser this method unlocks is intentionally left out of scope; it's a separate feature this PR makes possible.
🤖 Generated with Claude Code