feat: showcase text generation and thinking steps from suggested prompts #37

zimeg · 2025-12-18T00:09:43Z

Type of change

New feature

Summary

This PR showcases "chunks" in a chat stream. The example is meant to be quick to understand but perhaps not so meaningful. Related: slackapi/python-slack-sdk#1809

Requirements

I have ensured the changes I am contributing align with existing patterns and have tested and linted my code
I've read and agree to the Code of Conduct

jonigl · 2025-12-18T12:13:40Z

listeners/assistant/message.py

+        # The second example shows detailed thinking steps similar to tool calls
+        else:
+            streamer.append(
+                chunks=[
+                    MarkdownTextChunk(
+                        text="Hello.\nI have received the task. ",
+                    ),
+                    MarkdownTextChunk(
+                        text="This task appears manageable.\nThat is good.",
+                    ),
+                    TaskUpdateChunk(
+                        id="001",
+                        title="Understanding the task...",
+                        status="in_progress",
+                        details="- Identifying the goal\n- Identifying constraints",
+                    ),
+                    TaskUpdateChunk(
+                        id="002",
+                        title="Performing acrobatics...",
+                        status="pending",
+                    ),
+                ],
+            )
+            time.sleep(4)
+
+            streamer.append(
+                chunks=[
+                    TaskUpdateChunk(
+                        id="001",
+                        title="Understanding the task...",
+                        status="complete",
+                        details="\n- Pretending this was obvious",
+                        output="We'll continue to ramble now",
+                    ),
+                    TaskUpdateChunk(
+                        id="002",
+                        title="Performing acrobatics...",
+                        status="in_progress",
+                    ),
+                ],
+            )
+            time.sleep(4)


Nice! I've been experimenting with something similar on my own, I like this idea. Featuring what is the agent doing is great like (tools calls or thinking steps)

srtaalej · 2025-12-18T19:46:07Z

listeners/events/app_mentioned.py

        )

-        returned_message = call_llm([{"role": "user", "content": text}])
+        returned_message = call_llm(text)


zimeg · 2026-01-19T06:30:55Z

📸 Demo with current changes!

Timeline with tool calls and LLM call

dice.mov

Plan with mocked data

plan.mov

mwbrooks

✅ Super fun prompts that showcase the timeline and plan display styles alongside the new task and plan blocks. I think this is a good start to inspiring developers to integrate their own tool calls.

🧪 Local testing worked like a charm. I didn't find any issues.

📝 I left a few suggestions. Nothing is blocking, but I'd ask to consider switching the inverse logic (maybe I'm just a little slow in the ol'noodle and need it outlined simpler). Feel free to pump any-or-all to the future.

🚀 Please drop that draft label so we can smash the merge!

agent/tools/dice.py

mwbrooks · 2026-01-19T23:18:14Z

listeners/assistant/message.py

+        # This first example shows a generated text response for the provided prompt
+        if message["text"] != "Wonder a few deep thoughts.":


suggestion: It took me a few re-reads before I noticed the inverted logic (!=).

At first, I kept wondering why "Wonder a few deep thoughts." would be passed directly to the LLM, while the else statement handled the example.

Non-blocker, but could we flip this logic to be == "Wonder a few deep thoughts.": and have the else handle the general LLM response?

mwbrooks · 2026-01-19T23:19:58Z

listeners/assistant/message.py

+            streamer.append(
+                chunks=[
+                    MarkdownTextChunk(
+                        text="Hello.\nI have received the task. ",
+                    ),
+                    MarkdownTextChunk(
+                        text="This task appears manageable.\nThat is good.",
+                    ),
+                    TaskUpdateChunk(
+                        id="001",
+                        title="Understanding the task...",
+                        status="in_progress",
+                        details="- Identifying the goal\n- Identifying constraints",
+                    ),
+                    TaskUpdateChunk(
+                        id="002",
+                        title="Performing acrobatics...",
+                        status="pending",
+                    ),
+                ],
+            )
+            time.sleep(4)


praise: Love this example. It's dead simple and easy to grok. It's ripe to be altered, experimented, and hacked on for folks playing with the sample. 🙌🏻

mwbrooks · 2026-01-19T23:21:10Z

ai/tools/__init__.py

suggestion: I think it's reasonable to rename ai/ to agent/ if we want, but it can also happen in a later PR.

@mwbrooks Awesome callout! Let's make it in this PR as part of 6ab8795 with the addition of tools 🤖

listeners/assistant/message.py

srtaalej · 2026-01-20T19:02:12Z

agent/tools/dice.py

i really like this example 🤩

@srtaalej It makes for fun games! 🎲 ✨

srtaalej

Nice changes and its working for me ⭐ ⭐ ⭐

srtaalej · 2026-01-20T21:08:01Z

agent/llm_caller.py

-    messages.extend(messages_in_thread)
-    response = openai_client.responses.create(
-        model="gpt-4o-mini", input=messages, stream=True
+    streamer: ChatStream,


what do we think about moving the OpenAI client initialization to the module level instead of creating it inside call_llm() 🤔 that way we’re not re-initializing the client on every call
it’s probably fine either way for this dice-rolling use case, but it might be better to model best practices for users ❓

@srtaalej That might be nice to follow up with! At the moment I find the following error perhaps due to import order and the "dotenv" package of the main module though:

openai.OpenAIError: The api_key client option must be set either by passing api_key to the client or by setting the OPENAI_API_KEY environment variable

🗣️ ramble: We might also consider keeping it within a function to avoid global variables, but I do not know the best practices of modules, I admit!

Ah, i was getting the same error when i was working on the app. Ill try to figure out how we fixed it in case we run into it again but good work! ⭐ ⭐ ⭐

Co-authored-by: Michael Brooks <michael@michaelbrooks.ca>

zimeg

@jonigl @srtaalej @mwbrooks Thanks all for taking a look at these changes and the kind encouragement 💌

I made a few changes around code structure to make understanding easier I hope, but we might want to follow up with instructions around development with the CLI too.

Let's merge this after a bit more testing 🧪 ✨

agent/tools/dice.py

zimeg · 2026-01-20T23:23:00Z

ai/tools/__init__.py

@mwbrooks Awesome callout! Let's make it in this PR as part of 6ab8795 with the addition of tools 🤖

zimeg · 2026-01-20T23:23:35Z

agent/tools/dice.py

@srtaalej It makes for fun games! 🎲 ✨

zimeg · 2026-01-20T23:27:16Z

agent/llm_caller.py

-    messages.extend(messages_in_thread)
-    response = openai_client.responses.create(
-        model="gpt-4o-mini", input=messages, stream=True
+    streamer: ChatStream,


@srtaalej That might be nice to follow up with! At the moment I find the following error perhaps due to import order and the "dotenv" package of the main module though:

openai.OpenAIError: The api_key client option must be set either by passing api_key to the client or by setting the OPENAI_API_KEY environment variable

zimeg · 2026-01-20T23:29:26Z

agent/llm_caller.py

-    messages.extend(messages_in_thread)
-    response = openai_client.responses.create(
-        model="gpt-4o-mini", input=messages, stream=True
+    streamer: ChatStream,


🗣️ ramble: We might also consider keeping it within a function to avoid global variables, but I do not know the best practices of modules, I admit!

listeners/assistant/message.py

Co-authored-by: Michael Brooks <michael@michaelbrooks.ca>

zimeg · 2026-01-21T00:56:22Z

@mwbrooks Thanks for that last minute find too! I'll merge this now for upcoming testing and continued iterations 🚢 💨 🫡

zimeg added 2 commits December 17, 2025 15:45

feat: showcase text generation and thinking steps from suggested prompts

242297a

fix: append to details instead of overwrite past arguments

ea6c722

zimeg requested review from mwbrooks and srtaalej December 18, 2025 00:09

zimeg self-assigned this Dec 18, 2025

zimeg added the enhancement New feature or request label Dec 18, 2025

zimeg marked this pull request as draft December 18, 2025 00:09

jonigl reviewed Dec 18, 2025

View reviewed changes

srtaalej approved these changes Dec 18, 2025

View reviewed changes

feat: roll dice

4a5de84

mwbrooks approved these changes Jan 19, 2026

View reviewed changes

srtaalej self-requested a review January 20, 2026 18:59

srtaalej reviewed Jan 20, 2026

View reviewed changes

srtaalej approved these changes Jan 20, 2026

View reviewed changes

zimeg marked this pull request as ready for review January 20, 2026 19:52

srtaalej reviewed Jan 20, 2026

View reviewed changes

zimeg and others added 4 commits January 20, 2026 15:17

docs: improve comments to reason about sections of code

ac2816f

Co-authored-by: Michael Brooks <michael@michaelbrooks.ca>

refactor: move 'ai' features to an 'agent' directory

6ab8795

refactor: reverse the order of expected prompts for longform demo first

e2f6fa2

fix: match verbe tenses of actions and details in steps

7ddcac0

zimeg commented Jan 21, 2026

View reviewed changes

mwbrooks reviewed Jan 21, 2026

View reviewed changes

listeners/assistant/message.py Outdated Show resolved Hide resolved

docs: correct spellings of shocked wordage in example

d6af288

Co-authored-by: Michael Brooks <michael@michaelbrooks.ca>

zimeg merged commit b387ecf into feat-ai-apps-thinking-steps Jan 21, 2026
2 checks passed

zimeg deleted the zimeg-feat-chunks branch January 21, 2026 00:56

		# This first example shows a generated text response for the provided prompt
		if message["text"] != "Wonder a few deep thoughts.":

feat: showcase text generation and thinking steps from suggested prompts #37

feat: showcase text generation and thinking steps from suggested prompts #37

Conversation

zimeg commented Dec 18, 2025

Type of change

Summary

Requirements

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

zimeg commented Jan 19, 2026

Uh oh!

mwbrooks left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

srtaalej left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

zimeg left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

zimeg commented Jan 21, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants