Human-in-the-Loop Implementation #2021

mjschock · 2025-11-01T01:47:43Z

Resolves #636.

OPENAI_API_KEY="your_api_key_here" uv run python examples/agent_patterns/human_in_the_loop.py

================================================================================
Run interrupted - tool approval required
================================================================================
State saved to result.json
Loading state from result.json

Tool call details:
  Agent: Weather Assistant
  Tool: get_temperature
  Arguments: {"city":"Oakland"}

Do you approve this tool call? (y/n): y
✓ Approved: get_temperature

Resuming agent execution...

================================================================================
Final Output:
================================================================================
The weather in Oakland is sunny, and the temperature is 20°C.

OPENAI_API_KEY="your_api_key_here" uv run python examples/agent_patterns/human_in_the_loop_stream.py 

================================================================================
Human-in-the-loop: approval required for the following tool calls:
================================================================================

Tool call details:
  Agent: Weather Assistant
  Tool: get_temperature
  Arguments: {"city":"Oakland"}

Do you approve this tool call? (y/n): y
✓ Approved: get_temperature

Resuming agent execution...

================================================================================
Final Output:
================================================================================
The current weather in Oakland is sunny, with a temperature of 20°C.

Done!

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2025-11-01T18:19:41Z

src/agents/run.py

        previous_response_id: str | None,
        conversation_id: str | None,
        session: Session | None,
+        run_state: RunState[TContext] | None = None,
    ):
        if streamed_result.trace:
            streamed_result.trace.start(mark_as_current=True)


Prime server tracker when resuming streaming runs

When Runner.run_streamed resumes from a RunState with conversation_id or previous_response_id, _start_streaming constructs a _ServerConversationTracker but never seeds it with the prior model_responses that were already sent. Unlike the synchronous path, no call to track_server_items is made, so prepare_input treats every previously generated item as unsent and resubmits them to the server. This duplicates earlier messages and breaks server-side conversation threading when a run is resumed.

Useful? React with 👍 / 👎.

Fixed in a56ce0a. Added server conversation tracker priming at lines 1076-1079 to match the non-streaming implementation and prevent message duplication when resuming from RunState.

chatgpt-codex-connector · 2025-11-01T18:19:41Z

src/agents/run.py

        previous_response_id: str | None,
        conversation_id: str | None,
        session: Session | None,
+        run_state: RunState[TContext] | None = None,
    ):
        if streamed_result.trace:
            streamed_result.trace.start(mark_as_current=True)


Streaming resume ignores existing turn count

The streaming execution path always initializes current_turn = 0 when _start_streaming is called, even if a RunState with an existing _current_turn is supplied. The loop then increments from zero, so any turns completed before the interruption are ignored and the max_turns guard is reset. After each interruption, a resumed streaming run can exceed the user’s turn limit and misreport the current turn number.

Useful? React with 👍 / 👎.

This was already fixed in 74c50fd at line 914: current_turn=run_state._current_turn if run_state else 0. The turn counter is properly restored from the RunState.

…essage duplication When resuming a streaming run from RunState, the server conversation tracker was not being primed with previously sent model responses. This caused `prepare_input` to treat all previously generated items as unsent and resubmit them to the server, breaking conversation threading. **Issue**: Missing `track_server_items` call in streaming resumption path **Fix**: Added server conversation tracker priming logic in `_start_streaming` method (lines 1076-1079) to match the non-streaming path implementation (lines 553-556). The fix iterates through `run_state._model_responses` and calls `track_server_items(response)` to mark them as already sent to the server. **Impact**: Resolves message duplication when resuming interrupted streaming runs, ensuring proper conversation threading with server-side sessions. Fixes code review feedback from PR openai#2021 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>

seratch · 2025-11-04T02:33:20Z

Thanks for sending this patch!

I currently don't have the bandwidth to check this in depth, but one thing I wanted to mention is that, while implementing the sessions feature in openai-agents-js project, I found that the internals of runner need to take various HITL patterns into consideration. There might not be necessary to make those changes in this Python SDK, but sufficient testing for the sessions scenarios is worth doing.

mjschock · 2025-11-04T17:45:50Z

Thanks for sending this patch!

I currently don't have the bandwidth to check this in depth, but one thing I wanted to mention is that, while implementing the sessions feature in openai-agents-js project, I found that the internals of runner need to take various HITL patterns into consideration. There might not be necessary to make those changes in this Python SDK, but sufficient testing for the sessions scenarios is worth doing.

Happy to contribute! I added a couple examples using SQLiteSession and OpenAIConversationsSession and made sure they work:

OPENAI_API_KEY="your_api_key_here" uv run python examples/memory/memory_session_hitl_example.py 
=== Memory Session + HITL Example ===
Session id: :memory:
Enter a message to chat with the agent. Submit an empty line to exit.
The agent will ask for approval before using tools.

You: What cities does the Bay Bridge connect?
Assistant: The Bay Bridge connects San Francisco and Oakland in California.

You: What's the weather in those cities?

Agent HITL Assistant wants to call 'get_weather' with {"location":"San Francisco, CA"}. Approve? (y/n): y
Approved tool call.

Agent HITL Assistant wants to call 'get_weather' with {"location":"Oakland, CA"}. Approve? (y/n): y
Approved tool call.
Assistant: San Francisco is currently foggy with a temperature of 58°F. Oakland is sunny with a temperature of 72°F.

You:

OPENAI_API_KEY="your_api_key_here" uv run python examples/memory/openai_session_hitl_example.py 
=== OpenAI Session + HITL Example ===
Enter a message to chat with the agent. Submit an empty line to exit.
The agent will ask for approval before using tools.

You: What cities does the Bay Bridge connect?
Assistant: The Bay Bridge, officially known as the San Francisco–Oakland Bay Bridge, connects the cities of **San Francisco** and **Oakland** in California.

You: What's the weather in those cities?

Agent HITL Assistant wants to call 'get_weather' with {"location":"San Francisco, CA"}. Approve? (y/n): y
Approved tool call.

Agent HITL Assistant wants to call 'get_weather' with {"location":"Oakland, CA"}. Approve? (y/n): y
Approved tool call.
Assistant: San Francisco is currently foggy and 58°F, while Oakland is sunny and 72°F.

You:

I'm hoping that just about covers everything but lemme know if there are other areas I should make sure address.

…essage duplication When resuming a streaming run from RunState, the server conversation tracker was not being primed with previously sent model responses. This caused `prepare_input` to treat all previously generated items as unsent and resubmit them to the server, breaking conversation threading. **Issue**: Missing `track_server_items` call in streaming resumption path **Fix**: Added server conversation tracker priming logic in `_start_streaming` method (lines 1076-1079) to match the non-streaming path implementation (lines 553-556). The fix iterates through `run_state._model_responses` and calls `track_server_items(response)` to mark them as already sent to the server. **Impact**: Resolves message duplication when resuming interrupted streaming runs, ensuring proper conversation threading with server-side sessions. Fixes code review feedback from PR openai#2021 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>

mjschock · 2025-11-06T18:48:09Z

Thanks @seratch for adding it to the 0.6.x milestone! I'll make sure to address any feedback.

mjschock · 2025-11-08T19:52:35Z

Just a heads up. I noticed there are some discrepancies with the serialized RunState when interrupting in Python and resuming in TypeScript and visa versa. I'm addressing those this weekend and will push the update.

…essage duplication When resuming a streaming run from RunState, the server conversation tracker was not being primed with previously sent model responses. This caused `prepare_input` to treat all previously generated items as unsent and resubmit them to the server, breaking conversation threading. **Issue**: Missing `track_server_items` call in streaming resumption path **Fix**: Added server conversation tracker priming logic in `_start_streaming` method (lines 1076-1079) to match the non-streaming path implementation (lines 553-556). The fix iterates through `run_state._model_responses` and calls `track_server_items(response)` to mark them as already sent to the server. **Impact**: Resolves message duplication when resuming interrupted streaming runs, ensuring proper conversation threading with server-side sessions. Fixes code review feedback from PR openai#2021 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>

seratch · 2025-11-16T03:27:40Z

@codex Can you do reviews on the code changes in this PR with fresh eyes? Please focus on the potential issues for existing apps and edge case scenarios rather than general feedback.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2025-11-16T03:31:21Z

src/agents/run.py


            await AgentRunner._save_result_to_session(session, starting_input, [])


Avoid duplicating session history when resuming streaming runs

When Runner.run_streamed is resumed from a RunState, _start_streaming still calls _prepare_input_with_session and _save_result_to_session unconditionally (streamed_result.input is recomputed and the original prompt is saved again). In the non‑streaming path these calls are skipped when a RunState is supplied so that the session history isn’t duplicated. Re‑preparing and saving here means a resumed streaming run writes the same user message back into the session and the next turn will see duplicate user prompts, breaking conversations for apps that rely on Session or session_input_callback. Consider gating these calls with run_state is None as the synchronous path does.

Useful? React with 👍 / 👎.

👍 Fixed in 466c306

…essage duplication When resuming a streaming run from RunState, the server conversation tracker was not being primed with previously sent model responses. This caused `prepare_input` to treat all previously generated items as unsent and resubmit them to the server, breaking conversation threading. **Issue**: Missing `track_server_items` call in streaming resumption path **Fix**: Added server conversation tracker priming logic in `_start_streaming` method (lines 1076-1079) to match the non-streaming path implementation (lines 553-556). The fix iterates through `run_state._model_responses` and calls `track_server_items(response)` to mark them as already sent to the server. **Impact**: Resolves message duplication when resuming interrupted streaming runs, ensuring proper conversation threading with server-side sessions. Fixes code review feedback from PR openai#2021 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>

… to_input_item() for RunItem instances

…_input after resume

…or improved filtering of rehydrated items

…servation during resumption

…tate management

…ll items already persisted

…tes during resume scenarios

…oval are executed before shell tools

…r function and shell tools

…approved tool

…oceeding with new items

…napproved tools once

…processed tools by filtering based on call_ids

…l deserialization

…pproval requests

… duplicates during resume scenarios" This reverts commit 230225c398ec33b8b07cb6ad9fd8286f8288be01.

…g with all items already persisted" This reverts commit 5373de8e63197951c07f2af9a79a0382d9899eba.

… all items already persisted

…tion scenarios

mjschock · 2025-12-22T23:20:07Z

Thank you so much for maintaining this branch for a while! 🙇 I will start checking this in depth next week, but before that, I ran local codex reviews (/review) so many times and got these feedback from it. I've verified there is no definite duplicate in the list but some of them might be already mentioned in this pull request discussion. Hope this helps!

Also, from my experience for TS SDK, I do understand implementing this along with sessions compatibility is really complex. That said, it seems this PR has huge diff now and it could be challenging to review for me. If you see anything to simplify in this PR, refactoring the code would be appreciated as well.

* [P1] Server conversation tracker returns non-serializable run items — openai-agents-python/src/agents/run.py:456-492
  In _ServerConversationTracker.prepare_input the loop over generated_items now appends run_item.raw_item directly (lines 456‑491) instead of using run_item.to_input_item() as before. For run
  items whose raw_item is a Pydantic model (e.g., ResponseFunctionToolCall or ResponseOutputMessage), this returns a model instance rather than the dict payload the Responses API expects, so the
  next request will contain non‑JSON‑serializable objects and fail whenever a BaseModel output/tool call needs to be replayed to the server-managed conversation. The code should continue
  converting run items to input dictionaries before adding them to the request.

* [P1] Generated items are repeatedly resent to model — openai-agents-python/src/agents/run.py:431-492
  In _ServerConversationTracker.prepare_input generated items are appended to input_items but never recorded in sent_items as before, so subsequent calls will treat the same raw_item as unseen and
  resend it to the model/session. Any multi-turn or retried run that calls prepare_input more than once will duplicate previously sent tool outputs/messages, bloating history and potentially
  changing model behavior or causing server-side "duplicate item" errors. Add the sent-item bookkeeping when enqueuing generated items to avoid resending them.

* [P1] Server-managed conversation sends unnormalized tool outputs — openai-agents-python/src/agents/run.py:431-492
  When server-managed conversations are used (conversation_id/previous_response_id flows), _ServerConversationTracker.prepare_input now injects each generated run item’s raw_item directly into the
  next request. ToolCallOutputItem.to_input_item() used to strip fields the Responses API rejects (e.g., status, shell_output, protocol-only function_call_result names). Skipping this
  normalization means hosted tool outputs or function results with protocol fields are sent verbatim and will be rejected by the API, breaking any turn that includes local tool execution while
  using server-managed conversations.

* [P1] prepare_input ignores fingerprinted sent items — openai-agents-python/src/agents/run.py:431-490
  In _ServerConversationTracker.prepare_input we only skip generated items that match by object id or server-tracked ids, but we never consult sent_item_fingerprints even though prime_from_state
  and track_server_items populate it specifically to cover rehydrated copies. When resuming a server-managed conversation (or after deserializing state), the generated items are new objects with
  different ids, so prepare_input will re-send the entire prior history/tool outputs back to the Responses API, duplicating messages and tool outputs. We should also skip items whose serialized
  fingerprint is already in sent_item_fingerprints to avoid replaying previously delivered content after a resume.

* [P1] Interrupted runs lose model response for resumption — openai-agents-python/src/agents/run.py:1213-1244
  When a fresh run hits a NextStepInterruption, the returned RunResult is built without appending the current step’s model_response or setting _last_processed_response (lines 1213‑1244). Calling
  to_state() and attempting to resume a HITL flow then produces a RunState with empty _model_responses and no processed response, causing Runner.run to raise UserError("No model response found in
  previous state"). Any interruption on the first pass is therefore not resumable. The model response and processed response need to be captured on this path before returning.

* [P1] Duplicates already persisted session items when resuming — src/agents/run.py:4195-4203
  When saving conversation history, _save_result_to_session is intended to skip items that were already persisted in a prior partial save, but if _current_turn_persisted_item_count >=
  len(new_items) it sets new_run_items = list(new_items) rather than an empty list. In resumed HITL flows this resubmits items that were already saved earlier in the turn, leading to duplicated
  messages/tool outputs in the backing session (OpenAI Conversations or SQLite) and potentially “item already exists” errors. The early-return should skip saving when everything is already
  persisted instead of resaving the entire list.

* [P1] Approvals run after already executing other tools — openai-agents-python/src/agents/_run_impl.py:350-432
  In RunImpl.execute_tools_and_side_effects we start all tool executions with asyncio.gather before checking for ToolApprovalItem interruptions, so computer/shell/apply_patch calls in the same
  model response will still run even when a function tool needs approval. This defeats the HITL expectation that the turn pauses before any side effects occur; e.g., a response containing both a
  tool requiring approval and a computer action will still perform the computer action without approval. The interruption check should precede executing non-approved tool calls so that no side
  effects happen until the approval is granted.

* [P1] Pending approvals ignored when resuming interrupted turn — openai-agents-python/src/agents/_run_impl.py:678-754
  When resuming from a saved RunState, resolve_interrupted_turn executes the queued tools and appends their results to new_items, but it never checks whether any of those executions returned a
  ToolApprovalItem. If a tool still requires approval (e.g., the user didn’t approve it, or an on_approval callback leaves it pending), ShellAction/ApplyPatchAction/FunctionTool will return a
  ToolApprovalItem, yet this code proceeds to NextStepRunAgain without setting NextStepInterruption or populating interruptions. The caller receives no interruption and the next turn will filter
  out the approval items, so the HITL approval is silently dropped and the tool never runs. The interruption check needs to mirror the one in execute_tools_and_side_effects to return
  NextStepInterruption whenever any tool still needs approval.

* [P1] Resuming after approval replays already-run tools — openai-agents-python/src/agents/_run_impl.py:690-724
  When a turn is interrupted for tool approval, tools that did not require approval are executed before the interruption is raised. On resume, resolve_interrupted_turn runs all computer/shell/
  apply_patch calls again (execute_computer_actions, execute_shell_calls, etc.) without filtering out the ones already run, and then appends their outputs because object IDs differ. If a model
  response mixes an approval-gated function call with a shell/apply_patch action, resuming after approval will re-run that shell/patch a second time, duplicating side effects and history entries.
  These calls should be skipped on resume unless they were awaiting approval.

* [P1] Drop providerData when deserializing interrupts MCP approvals — openai-agents-python/src/agents/run_state.py:1471-1484
  RunState deserialization strips providerData from stored items via _normalize_field_names, so a serialized ToolApprovalItem for a hosted MCP tool loses the providerData.type marker required to
  detect MCP approval requests when resuming. As a result, if a run is interrupted for MCP approval, saved to JSON, and later restored, resolve_interrupted_turn never finds those pending MCP
  approvals and they will not be executed or surfaced to the user, blocking continuation of that flow.

I'm happy to contribute and I really appreciate you sending the edge cases my way. I've created tests that capture each failing edge case and the corresponding fixes. I'll spend some time this week to see if I can reduce the number of changes while maintaining the test cases. Less is more. =)

mjschock · 2025-12-22T23:26:15Z

@codex please take a fresh look at the full set of changes and do a thorough review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

chatgpt-codex-connector · 2025-12-22T23:32:25Z

src/agents/run_state.py

+        # Add additional fields based on item type
+        if hasattr(item, "output"):
+            result["output"] = str(item.output)


Preserve tool outputs when serializing run state

RunState serialization coerces every item’s output to a string (result["output"] = str(item.output)), so any structured tool result (e.g., a dict/list from a function tool or shell output) is irreversibly stringified when saving to JSON. _deserialize_items later feeds this string back into ToolCallOutputItem, meaning resumed runs or resumed session persistence no longer see the original structured output. HITL flows that persist state to disk will thus lose typed tool outputs after resumption. Avoid stringifying the output so it round-trips with its original type.

Useful? React with 👍 / 👎.

… run state serialization

…tructured types and ensure compatibility

mjschock · 2025-12-23T00:45:02Z

@codex review

chatgpt-codex-connector · 2025-12-23T00:50:54Z

Codex Review: Didn't find any major issues. Keep it up!

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

seratch · 2025-12-23T00:52:13Z

@mjschock Thank you so much for taking the time to work on this. Since yesterday, I've been working on the code based on your branch here. There are still some things to polish, and there are additional patterns that need to be supported (e.g., local shell tool migration and perhaps some others).

This enhancement is a fairly large change, and as a maintainer, I need to ensure the final quality. If you're okay with it, would it be all right if I take over your changes in a new pull request and continue collaborating with you on the remaining work? Of course, I'll credit your contribution using Co-authored-by, so your work will be fully preserved in the new PR. I'd really appreciate your understanding.

mjschock · 2025-12-23T00:56:09Z

@mjschock Thank you so much for taking the time to work on this. Since yesterday, I've been working on the code based on your branch here. There are still some things to polish, and there are additional patterns that need to be supported (e.g., local shell tool migration and perhaps some others).

This enhancement is a fairly large change, and as a maintainer, I need to ensure the final quality. If you're okay with it, would it be all right if I take over your changes in a new pull request and continue collaborating with you on the remaining work? Of course, I'll credit your contribution using Co-authored-by, so your work will be fully preserved in the new PR. I'd really appreciate your understanding.

@seratch yep, np

mjschock mentioned this pull request Nov 1, 2025

Human-In-The-Loop Architecture should be implemented on top priority! #636

Open

mjschock marked this pull request as ready for review November 1, 2025 18:16

chatgpt-codex-connector bot reviewed Nov 1, 2025

View reviewed changes

seratch added enhancement New feature or request feature:core labels Nov 4, 2025

mjschock force-pushed the main branch from b262f66 to 3d52667 Compare November 4, 2025 17:47

mjschock force-pushed the main branch from 3d52667 to f30310e Compare November 4, 2025 17:55

mjschock force-pushed the main branch from f30310e to 110dcbd Compare November 5, 2025 02:15

seratch added this to the 0.6.x milestone Nov 5, 2025

mjschock force-pushed the main branch from 110dcbd to b38b2f7 Compare November 6, 2025 18:45

mjschock force-pushed the main branch from 2bc0ce0 to 9f37210 Compare November 14, 2025 02:12

mjschock force-pushed the main branch from 9f37210 to 1947718 Compare November 14, 2025 16:54

mjschock force-pushed the main branch from 1947718 to def65ac Compare November 16, 2025 01:42

chatgpt-codex-connector bot reviewed Nov 16, 2025

View reviewed changes

mjschock force-pushed the main branch from 66df149 to 9ca00a8 Compare November 17, 2025 23:43

seratch removed this from the 0.6.x milestone Nov 18, 2025

mjschock added 20 commits December 22, 2025 10:36

fix: ensure consistent normalization of input items by always calling…

03cd28a

… to_input_item() for RunItem instances

test: add test to verify filtering of items by fingerprint in prepare…

b4c5f74

…_input after resume

feat: add original_input_fingerprints to _ServerConversationTracker f…

5a36beb

…or improved filtering of rehydrated items

test: add test for interrupted run state to ensure model response pre…

71d0b53

…servation during resumption

fix: add processed_response to RunImpl and AgentRunner for improved s…

7b4f181

…tate management

test: add test to verify no duplicates are saved when resuming with a…

ff0b7df

…ll items already persisted

fix: refine turn management and input saving logic to prevent duplica…

275a02a

…tes during resume scenarios

test: add approval check test to ensure function tools requiring appr…

183d3c5

…oval are executed before shell tools

fix: reorganize tool execution order to prioritize approval checks fo…

2a71b86

…r function and shell tools

test: add test to verify interruption is raised when resuming with un…

a0635a4

…approved tool

fix: implement tool approval checks to handle interruptions before pr…

7ea1cf3

…oceeding with new items

test: add test to verify that resuming after approval only executes u…

880888f

…napproved tools once

fix: enhance tool execution logic to prevent re-execution of already …

ac49221

…processed tools by filtering based on call_ids

test: add test to ensure providerData is preserved during MCP approva…

c766e13

…l deserialization

fix: preserve and restore providerData during normalization for MCP a…

980f49e

…pproval requests

Revert "fix: refine turn management and input saving logic to prevent…

e3efb5d

… duplicates during resume scenarios" This reverts commit 230225c398ec33b8b07cb6ad9fd8286f8288be01.

Revert "test: add test to verify no duplicates are saved when resumin…

7d9afa0

…g with all items already persisted" This reverts commit 5373de8e63197951c07f2af9a79a0382d9899eba.

fix: address lint issues after rebase

608abe3

test: add test to verify no duplicates are created when resuming with…

e4dec73

… all items already persisted

fix: refine logic to prevent duplicate item persistence during resump…

28c008e

…tion scenarios

mjschock force-pushed the main branch from 19b5b89 to 28c008e Compare December 22, 2025 23:08

chatgpt-codex-connector bot reviewed Dec 22, 2025

View reviewed changes

mjschock added 2 commits December 22, 2025 15:44

test: add test to ensure structured tool outputs are preserved during…

4248b87

… run state serialization

fix: enhance output handling in run state serialization to preserve s…

8341547

…tructured types and ensure compatibility


		await AgentRunner._save_result_to_session(session, starting_input, [])

Human-in-the-Loop Implementation #2021

Are you sure you want to change the base?

Human-in-the-Loop Implementation #2021

Conversation

mjschock commented Nov 1, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Nov 1, 2025

Choose a reason for hiding this comment

Uh oh!

mjschock Nov 1, 2025

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector bot Nov 1, 2025

Choose a reason for hiding this comment

Uh oh!

mjschock Nov 1, 2025

Choose a reason for hiding this comment

Uh oh!

seratch commented Nov 4, 2025

Uh oh!

mjschock commented Nov 4, 2025

Uh oh!

mjschock commented Nov 6, 2025

Uh oh!

mjschock commented Nov 8, 2025

Uh oh!

seratch commented Nov 16, 2025

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Nov 16, 2025

Choose a reason for hiding this comment

Uh oh!

mjschock Nov 17, 2025

Choose a reason for hiding this comment

Uh oh!

mjschock commented Dec 22, 2025

Uh oh!

mjschock commented Dec 22, 2025

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Dec 22, 2025

Choose a reason for hiding this comment

Uh oh!

mjschock commented Dec 23, 2025

Uh oh!

chatgpt-codex-connector bot commented Dec 23, 2025

Uh oh!

seratch commented Dec 23, 2025

Uh oh!

mjschock commented Dec 23, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

mjschock commented Nov 1, 2025 •

edited

Loading