fix(delegate): salvage #21933 JSON-string batch + diagnostic logging by kshitijk4poor · Pull Request #22436 · NousResearch/hermes-agent

kshitijk4poor · 2026-05-09T09:17:33Z

Summary

Salvage of #21966 (Bartok9) plus the model_tools.py diagnostic logging from #22092 (uzunkuyruk), addressing #21933 — delegate_task batch mode silently failing when open-weight models (Qwen, DeepSeek, GLM) emit the tasks array as a JSON-encoded string instead of a native array.

Also closes the duplicate PRs #21957, #21966, #22092.

Why a salvage

Three PRs targeted #21933:

PR	Author	Verdict
#21957	liuhao1024	Right idea, but mixed in unrelated WhatsApp `npm install` timeout changes and a `PATCH_SCHEMA` edit that incorrectly marks mode-specific params as `required`
#21966	Bartok9	Clean, focused, with tests. Picked.
#22092	uzunkuyruk	Adds useful diagnostic logging in `model_tools.py`, but its `delegate_task` recovery path targets a wrapped-list form (`tasks=["[json]"]`) that never reaches `delegate_task` — the function is in `_AGENT_LOOP_TOOLS` and bypasses `coerce_tool_args` entirely (dispatched directly from `run_agent.py` via `_dispatch_delegate_task` with raw `function_args`). Logging changes salvaged; the `delegate_tool.py` recovery dropped.

What this PR does

Commit 1 — fix(delegate): accept JSON string batch tasks (Bartok9, cherry-picked from #21966)

Adds a _recover_tasks_from_json_string() helper at the top of delegate_task() that:

accepts tasks arriving as a raw JSON-encoded string and parses it into a list (the actual production failure mode for [Bug]: delegate_task batch mode fails silently when model emits tasks as JSON string #21933)
returns a structured tool_error for malformed JSON, non-array JSON, or empty strings
adds per-element type validation rejecting non-object list entries with a clear message

Commit 2 — fix(model_tools): log warnings for failed JSON-array coercion (uzunkuyruk, salvaged from #22092)

Adds WARNING-level logs in _coerce_json and coerce_tool_args so that when any other tool (the ones that DO go through handle_function_call → coerce_tool_args) gets a stringified array argument that fails to parse, it's no longer silent. The delegate_task-specific recovery from #22092 was dropped — the actual fix is the previous commit.

Tests

3 new tests from #21966 (Bartok9):

test_batch_mode_accepts_json_string_tasks — JSON-encoded array string runs the batch
test_batch_mode_rejects_non_object_tasks — non-object list elements yield clear error
test_batch_mode_rejects_malformed_json_string_tasks — malformed JSON yields clear error

All pass. Full tests/tools/test_delegate.py (123 tests) and tests/run_agent/test_tool_arg_coercion.py (61 tests) green locally on this branch.

E2E verification

Confirmed the actual production failure mode, with both fixes applied:

Test 1: tasks = '[{"goal":"task1"},{"goal":"task2"}]'  (raw string, the bug)
  → batch executes 2 tasks ✓

Test 2: tasks = '[{"goal": "bad}'                       (malformed JSON)
  → "tasks must be a JSON array of task objects; received a string that
     could not be parsed as JSON (Unterminated string starting at)." ✓

Test 3: tasks = '{"goal": "x"}'                         (non-array JSON)
  → "tasks must be a JSON array of task objects; parsed dict instead." ✓

Test 4: tasks = [{"goal":"task1"},{"goal":"task2"}]     (native list, regression)
  → batch executes 2 tasks ✓

Test 5: tasks = ["not an object"]                       (bad list element)
  → "Task 0 must be an object, got str." ✓

Test 6: goal = "do something"                           (single-task mode)
  → reaches dispatch, tasks unaffected ✓

Test 7: model_tools._coerce_json('[{"bad json}', list)  (logging)
  → WARNING: "coerce_tool_args: failed to parse string as JSON for
     expected type list: Unterminated string starting at..." ✓

I also empirically verified that #22092's standalone delegate_task recovery does NOT fix the bug — running it alone, Test 1 still returns "Provide either 'goal' (single task) or 'tasks' (batch)." because coerce_tool_args never wraps the string into the form the PR's recovery expects.

Credit

fix(delegate): accept JSON string batch tasks #21966 by @Bartok9 — picked as the salvage base; their commit is preserved with original authorship.
fix(delegate-task): recover batch tasks from JSON-encoded string coer… #22092 by @uzunkuyruk — model_tools.py diagnostic logging cherry-picked with Author: preserved.
fix(delegate): normalize tasks-as-string in delegate_task for batch mode #21957 by @liuhao1024 — same intent at the delegate_task boundary; not picked due to scope creep, but the upstream issue diagnosis is valuable. Closing with credit.

Closes

Fixes [Bug]: delegate_task batch mode fails silently when model emits tasks as JSON string #21933
Closes fix(delegate): normalize tasks-as-string in delegate_task for batch mode #21957
Closes fix(delegate): accept JSON string batch tasks #21966 (via merge of cherry-picked commit)
Closes fix(delegate-task): recover batch tasks from JSON-encoded string coer… #22092 (via merge of cherry-picked logging commit)

Notes for reviewer

This salvage depends on #22434 (chore(release): add uzunkuyruk to AUTHOR_MAP) being merged first, since contributor_audit.py requires the email→login mapping for egitimviscara@gmail.com before this branch can land cleanly.

Merge with --rebase to preserve per-commit authorship.

Recover delegate_task batch inputs when open-weight models emit tasks as a JSON-encoded array string, and return clear errors for malformed task lists. Co-authored-by: Cursor <cursoragent@cursor.com>

When _coerce_json fails to parse a string as JSON or parses to the wrong type, log a clear WARNING instead of silently returning the original value. When coerce_tool_args wraps a bare string into a single-element list AND the string looks like a JSON array (starts with '['), warn that the model likely emitted a JSON-encoded string instead of a native array. This improves diagnostics for the open-weight model output drift described in #21933 (JSON-array-as-string), as well as any other tool whose array-typed argument arrives stringified through handle_function_call. Note: delegate_task does NOT go through coerce_tool_args (it is in _AGENT_LOOP_TOOLS and dispatched directly from run_agent.py with raw function_args from json.loads). The actual delegate_task fix for #21933 is the previous commit. These logging changes apply to all other array-typed arguments coerced via the shared pipeline. Salvaged from PR #22092.

github-actions · 2026-05-09T09:20:08Z

🔎 Lint report: `salvage/delegate-task-json-string-21933` vs `origin/main`

ruff

Total: 0 on HEAD, 0 on base (➖ 0)

🆕 New issues: none

✅ Fixed issues: none

Unchanged: 0 pre-existing issues carried over.

ty (type checker)

Total: 7875 on HEAD, 7854 on base (🆕 +21)

🆕 New issues (45):

Rule	Count
`invalid-argument-type`	34
`unresolved-attribute`	6
`unsupported-operator`	4
`not-subscriptable`	1

First entries

cli.py:8147: [invalid-argument-type] invalid-argument-type: Argument to function `estimate_usage_cost` is incorrect: Expected `str`, found `str | Unknown | dict[Unknown | str, Unknown | str | dict[str, str]] | int | dict[Unknown, Unknown]`
run_agent.py:8525: [invalid-argument-type] invalid-argument-type: Argument to function `get_transport` is incorrect: Expected `str`, found `str | Unknown | dict[Unknown | str, Unknown | str | dict[str, str]] | int | dict[Unknown, Unknown]`
run_agent.py:3075: [invalid-argument-type] invalid-argument-type: Argument to function `get_provider_stale_timeout` is incorrect: Expected `str`, found `str | Unknown | dict[Unknown | str, Unknown | str | dict[str, str]] | int | dict[Unknown, Unknown]`
run_agent.py:2562: [invalid-argument-type] invalid-argument-type: Argument to function `build_anthropic_client` is incorrect: Expected `str`, found `(Unknown & ~AlwaysFalsy) | (str & ~AlwaysFalsy) | (dict[str, str] & ~AlwaysFalsy) | ... omitted 4 union elements`
run_agent.py:8257: [invalid-argument-type] invalid-argument-type: Argument to bound method `ContextCompressor.update_model` is incorrect: Expected `int`, found `str | Unknown | dict[Unknown | str, Unknown | str | dict[str, str]] | int | dict[Unknown, Unknown]`
run_agent.py:5439: [unresolved-attribute] unresolved-attribute: Attribute `split` is not defined on `dict[Unknown | str, Unknown | str | dict[str, str]]`, `int`, `dict[Unknown, Unknown]` in union `str | Unknown | dict[Unknown | str, Unknown | str | dict[str, str]] | int | dict[Unknown, Unknown]`
run_agent.py:12286: [invalid-argument-type] invalid-argument-type: Argument to function `normalize_usage` is incorrect: Expected `str | None`, found `str | Unknown | dict[Unknown | str, Unknown | str | dict[str, str]] | int | dict[Unknown, Unknown]`
run_agent.py:2613: [invalid-argument-type] invalid-argument-type: Argument to function `get_model_context_length` is incorrect: Expected `str`, found `str | dict[str, str] | Any | ... omitted 3 union elements`
run_agent.py:12331: [invalid-argument-type] invalid-argument-type: Argument to function `estimate_usage_cost` is incorrect: Expected `str`, found `str | Unknown | dict[Unknown | str, Unknown | str | dict[str, str]] | int | dict[Unknown, Unknown]`
tests/run_agent/test_provider_attribution_headers.py:90: [unresolved-attribute] unresolved-attribute: Attribute `startswith` is not defined on `dict[str, str]` in union `Unknown | str | dict[str, str]`
tests/tools/test_delegate.py:196: [invalid-argument-type] invalid-argument-type: Argument to function `delegate_task` is incorrect: Expected `list[dict[str, Any]] | None`, found `str`
run_agent.py:3851: [invalid-argument-type] invalid-argument-type: Argument to `AIAgent.__init__` is incorrect: Expected `str`, found `str | Unknown | dict[Unknown | str, Unknown | str | dict[str, str]] | int | dict[Unknown, Unknown]`
tests/run_agent/test_provider_attribution_headers.py:155: [unsupported-operator] unsupported-operator: Operator `not in` is not supported between objects of type `Literal["X-OpenRouter-Cache"]` and `Unknown | str | dict[str, str] | ... omitted 3 union elements`
run_agent.py:2473: [invalid-argument-type] invalid-argument-type: Argument to function `ensure_lmstudio_model_loaded` is incorrect: Expected `str`, found `str | Unknown | dict[Unknown | str, Unknown | str | dict[str, str]] | int | dict[Unknown, Unknown]`
run_agent.py:6822: [invalid-argument-type] invalid-argument-type: Argument to function `build_anthropic_client` is incorrect: Expected `str`, found `str | dict[Unknown, Unknown] | Any | ... omitted 3 union elements`
run_agent.py:4988: [invalid-argument-type] invalid-argument-type: Argument to function `parse_rate_limit_headers` is incorrect: Expected `str`, found `str | Unknown | dict[Unknown | str, Unknown | str | dict[str, str]] | int | dict[Unknown, Unknown]`
run_agent.py:8341: [invalid-argument-type] invalid-argument-type: Argument to function `get_provider_request_timeout` is incorrect: Expected `str | None`, found `str | Unknown | dict[Unknown | str, Unknown | str | dict[str, str]] | int | dict[Unknown, Unknown]`
run_agent.py:4440: [invalid-argument-type] invalid-argument-type: Argument to function `save_trajectory` is incorrect: Expected `str`, found `str | Unknown | dict[Unknown | str, Unknown | str | dict[str, str]] | int | dict[Unknown, Unknown]`
run_agent.py:5439: [unsupported-operator] unsupported-operator: Operator `in` is not supported between objects of type `Literal["/"]` and `str | Unknown | dict[Unknown | str, Unknown | str | dict[str, str]] | int | dict[Unknown, Unknown]`
tests/agent/test_codex_cloudflare_headers.py:181: [unsupported-operator] unsupported-operator: Operator `in` is not supported between objects of type `Literal["originator"]` and `(Unknown & ~AlwaysFalsy) | (str & ~AlwaysFalsy) | (dict[str, str] & ~AlwaysFalsy) | ... omitted 3 union elements`
run_agent.py:9082: [invalid-argument-type] invalid-argument-type: Argument to function `lmstudio_model_reasoning_options` is incorrect: Expected `str`, found `str | Unknown | dict[Unknown | str, Unknown | str | dict[str, str]] | int | dict[Unknown, Unknown]`
run_agent.py:9361: [unresolved-attribute] unresolved-attribute: Attribute `lower` is not defined on `dict[Unknown | str, Unknown | str | dict[str, str]] & ~AlwaysFalsy`, `int & ~AlwaysFalsy`, `dict[Unknown, Unknown] & ~AlwaysFalsy` in union `(str & ~AlwaysFalsy) | (Unknown & ~AlwaysFalsy) | (dict[Unknown | str, Unknown | str | dict[str, str]] & ~AlwaysFalsy) | ... omitted 3 union elements`
run_agent.py:12304: [invalid-argument-type] invalid-argument-type: Argument to function `save_context_length` is incorrect: Expected `str`, found `str | Unknown | dict[Unknown | str, Unknown | str | dict[str, str]] | int | dict[Unknown, Unknown]`
tests/tools/test_delegate.py:208: [invalid-argument-type] invalid-argument-type: Argument to function `delegate_task` is incorrect: Expected `list[dict[str, Any]] | None`, found `list[dict[str, Any] | str]`
tests/tools/test_delegate.py:220: [invalid-argument-type] invalid-argument-type: Argument to function `delegate_task` is incorrect: Expected `list[dict[str, Any]] | None`, found `Literal["[{\"goal\": \"bad}"]`
... and 20 more

✅ Fixed issues (40):

Rule	Count
`invalid-argument-type`	34
`unresolved-attribute`	5
`unsupported-operator`	1

First entries

run_agent.py:7268: [invalid-argument-type] invalid-argument-type: Argument to function `get_provider_request_timeout` is incorrect: Expected `str | None`, found `str | Unknown | dict[Unknown, Unknown] | int | dict[Unknown | str, Unknown | str | dict[str, str]]`
run_agent.py:10839: [invalid-argument-type] invalid-argument-type: Argument to function `_fixed_temperature_for_model` is incorrect: Expected `str | None`, found `str | Unknown | dict[Unknown, Unknown] | int | dict[Unknown | str, Unknown | str | dict[str, str]]`
run_agent.py:2613: [invalid-argument-type] invalid-argument-type: Argument to function `get_model_context_length` is incorrect: Expected `str`, found `str | dict[str, str] | Any | ... omitted 4 union elements`
run_agent.py:11590: [invalid-argument-type] invalid-argument-type: Argument to function `apply_anthropic_cache_control` is incorrect: Expected `bool`, found `int | Divergent | Unknown | ... omitted 3 union elements`
run_agent.py:5439: [unsupported-operator] unsupported-operator: Operator `in` is not supported between objects of type `Literal["/"]` and `str | Unknown | dict[Unknown, Unknown] | int | dict[Unknown | str, Unknown | str | dict[str, str]]`
run_agent.py:3075: [invalid-argument-type] invalid-argument-type: Argument to function `get_provider_stale_timeout` is incorrect: Expected `str`, found `str | Unknown | dict[Unknown, Unknown] | int | dict[Unknown | str, Unknown | str | dict[str, str]]`
run_agent.py:3851: [invalid-argument-type] invalid-argument-type: Argument to `AIAgent.__init__` is incorrect: Expected `str`, found `str | Unknown | dict[Unknown, Unknown] | int | dict[Unknown | str, Unknown | str | dict[str, str]]`
run_agent.py:8257: [invalid-argument-type] invalid-argument-type: Argument to bound method `ContextCompressor.update_model` is incorrect: Expected `int`, found `Divergent | Unknown | str | ... omitted 3 union elements`
run_agent.py:13272: [invalid-argument-type] invalid-argument-type: Argument to bound method `ContextCompressor.update_model` is incorrect: Expected `str`, found `str | Unknown | dict[Unknown, Unknown] | int | dict[Unknown | str, Unknown | str | dict[str, str]]`
run_agent.py:8341: [invalid-argument-type] invalid-argument-type: Argument to function `get_provider_request_timeout` is incorrect: Expected `str`, found `Divergent | Unknown | str | ... omitted 3 union elements`
run_agent.py:5439: [unresolved-attribute] unresolved-attribute: Attribute `split` is not defined on `dict[Unknown, Unknown]`, `int`, `dict[Unknown | str, Unknown | str | dict[str, str]]` in union `str | Unknown | dict[Unknown, Unknown] | int | dict[Unknown | str, Unknown | str | dict[str, str]]`
run_agent.py:8340: [invalid-argument-type] invalid-argument-type: Argument to function `build_anthropic_client` is incorrect: Expected `str`, found `Divergent | Unknown | str | ... omitted 3 union elements`
run_agent.py:9082: [invalid-argument-type] invalid-argument-type: Argument to function `lmstudio_model_reasoning_options` is incorrect: Expected `str`, found `str | Unknown | dict[Unknown, Unknown] | int | dict[Unknown | str, Unknown | str | dict[str, str]]`
run_agent.py:3075: [invalid-argument-type] invalid-argument-type: Argument to function `get_provider_stale_timeout` is incorrect: Expected `str | None`, found `str | Unknown | dict[Unknown, Unknown] | int | dict[Unknown | str, Unknown | str | dict[str, str]]`
tests/agent/test_codex_cloudflare_headers.py:163: [unresolved-attribute] unresolved-attribute: Attribute `get` is not defined on `str & ~AlwaysFalsy` in union `(Unknown & ~AlwaysFalsy) | (str & ~AlwaysFalsy) | (dict[str, str] & ~AlwaysFalsy) | ... omitted 3 union elements`
run_agent.py:12304: [invalid-argument-type] invalid-argument-type: Argument to function `save_context_length` is incorrect: Expected `str`, found `str | Unknown | dict[Unknown, Unknown] | int | dict[Unknown | str, Unknown | str | dict[str, str]]`
run_agent.py:12286: [invalid-argument-type] invalid-argument-type: Argument to function `normalize_usage` is incorrect: Expected `str | None`, found `str | Unknown | dict[Unknown, Unknown] | int | dict[Unknown | str, Unknown | str | dict[str, str]]`
run_agent.py:2562: [invalid-argument-type] invalid-argument-type: Argument to function `build_anthropic_client` is incorrect: Expected `str`, found `(Unknown & ~AlwaysFalsy) | (str & ~AlwaysFalsy) | (dict[str, str] & ~AlwaysFalsy) | ... omitted 5 union elements`
cli.py:8147: [invalid-argument-type] invalid-argument-type: Argument to function `estimate_usage_cost` is incorrect: Expected `str`, found `str | Unknown | dict[Unknown, Unknown] | int | dict[Unknown | str, Unknown | str | dict[str, str]]`
run_agent.py:12333: [invalid-argument-type] invalid-argument-type: Argument to function `estimate_usage_cost` is incorrect: Expected `str | None`, found `str | Unknown | dict[Unknown, Unknown] | int | dict[Unknown | str, Unknown | str | dict[str, str]]`
run_agent.py:8936: [invalid-argument-type] invalid-argument-type: Argument to function `get_provider_profile` is incorrect: Expected `str`, found `str | Unknown | dict[Unknown, Unknown] | int | dict[Unknown | str, Unknown | str | dict[str, str]]`
run_agent.py:12818: [invalid-argument-type] invalid-argument-type: Argument to function `_is_oauth_token` is incorrect: Expected `str`, found `str | dict[Unknown | str, Unknown | str | dict[str, str]] | Any | ... omitted 5 union elements`
run_agent.py:2565: [invalid-argument-type] invalid-argument-type: Argument to function `_is_oauth_token` is incorrect: Expected `str`, found `(Unknown & ~AlwaysFalsy) | (str & ~AlwaysFalsy) | (dict[str, str] & ~AlwaysFalsy) | ... omitted 5 union elements`
run_agent.py:2330: [invalid-argument-type] invalid-argument-type: Argument to function `query_ollama_num_ctx` is incorrect: Expected `str`, found `(str & ~AlwaysFalsy) | (dict[str, str] & ~AlwaysFalsy) | (Any & ~AlwaysFalsy) | ... omitted 5 union elements`
run_agent.py:8525: [invalid-argument-type] invalid-argument-type: Argument to function `get_transport` is incorrect: Expected `str`, found `str | Unknown | dict[Unknown, Unknown] | int | dict[Unknown | str, Unknown | str | dict[str, str]]`
... and 15 more

Unchanged: 4127 pre-existing issues carried over.

Diagnostics are surfaced as warnings — this check never fails the build.

Bartok9 and others added 2 commits May 9, 2026 14:43

fix(delegate): accept JSON string batch tasks

b2ce706

Recover delegate_task batch inputs when open-weight models emit tasks as a JSON-encoded array string, and return clear errors for malformed task lists. Co-authored-by: Cursor <cursoragent@cursor.com>

kshitijk4poor merged commit 7330183 into main May 9, 2026
13 of 15 checks passed

kshitijk4poor deleted the salvage/delegate-task-json-string-21933 branch May 9, 2026 09:18

This was referenced May 9, 2026

fix(delegate): accept JSON string batch tasks #21966

Closed

fix(delegate-task): recover batch tasks from JSON-encoded string coer… #22092

Closed

kshitijk4poor mentioned this pull request May 9, 2026

fix(delegate): normalize tasks-as-string in delegate_task for batch mode #21957

Closed

alt-glitch added type/bug Something isn't working P2 Medium — degraded but workaround exists comp/agent Core agent loop, run_agent.py, prompt builder tool/delegate Subagent delegation labels May 9, 2026

github-actions Bot mentioned this pull request May 17, 2026

chore: bump NousResearch/hermes-agent version from v2026.5.7 to v2026.5.16 Docker-Hub-sirmark/docker-hermes-agent#6

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(delegate): salvage #21933 JSON-string batch + diagnostic logging#22436

fix(delegate): salvage #21933 JSON-string batch + diagnostic logging#22436
kshitijk4poor merged 2 commits into
mainfrom
salvage/delegate-task-json-string-21933

kshitijk4poor commented May 9, 2026

Uh oh!

Uh oh!

github-actions Bot commented May 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

kshitijk4poor commented May 9, 2026

Summary

Why a salvage

What this PR does

Tests

E2E verification

Credit

Closes

Notes for reviewer

Uh oh!

Uh oh!

github-actions Bot commented May 9, 2026

🔎 Lint report: salvage/delegate-task-json-string-21933 vs origin/main

ruff

ty (type checker)

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

🔎 Lint report: `salvage/delegate-task-json-string-21933` vs `origin/main`