fix: flatten rich message cards into text for webchat frontend

[TheDarkSkyXD]

Summary

• The webchat frontend doesn't support rich embeds (cards), so card content (titles, descriptions, fields) was silently lost when the bot replied with RichMessage variants
• Flattens card content into plain text using OutboundResponse::text_from_cards() across all webchat paths: SSE event forwarding (main.rs), broadcast (webchat.rs), conversation logging (reply.rs), and cancelled-history extraction (channel_history.rs)
• No behavioral change for Discord/Slack which render cards natively

Test plan

• Send a message that triggers a card-based response via the webchat UI — verify the full card content appears as formatted text
• Verify conversation history reload shows the flattened card content (not just the short text field)
• Confirm Discord/Slack responses still render rich embeds normally (no regression)
• Test edge cases: cards with no title, cards with no description, empty card list, message with text + cards, message with only cards

🤖 Generated with Claude Code

[coderabbitai]

Walkthrough

Adds handling of cards in outbound rich messages: cards are deserialized/collected, converted to plain text via OutboundResponse::text_from_cards, and combined with message text using OutboundResponse::text_with_cards across message emission, logging, and history extraction paths.

Changes

Cohort / File(s)	Summary
Channel history extraction src/agent/channel_history.rs	extract_reply_content_from_cancelled_history now reads reply tool call cards, attempts to deserialize to Vec<crate::Card> (warns and defaults on failure), and returns OutboundResponse::text_with_cards(...) instead of plain content.
SSE / main event forwarding src/main.rs	forward_sse_event adds a dedicated RichMessage { text, cards, .. } arm and computes full_text via OutboundResponse::text_with_cards(text, cards) before emitting ApiEvent::OutboundMessage.
WebChat broadcasting src/messaging/webchat.rs	WebChatAdapter::broadcast now handles OutboundResponse::RichMessage { text, cards, .. } by flattening cards into the sent text using text_with_cards.
Reply tool response & logging src/tools/reply.rs	ReplyTool::call extracts args.cards / args.interactive_elements into locals and uses OutboundResponse::text_with_cards to produce logged_content (so logs include flattened card text when present).
OutboundResponse helper src/lib.rs	Added public OutboundResponse::text_with_cards(text: &str, cards: &[Card]) -> String which derives text from cards and merges it with provided text using defined precedence/formatting.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately summarizes the main change: flattening rich message cards into text for the webchat frontend, which is the core objective across all modified files.
Description check	✅ Passed	The description is well-structured and directly relates to the changeset, explaining the problem (lost card content), the solution (flattening cards), affected code paths, and test coverage.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🧹 Nitpick comments (1)

src/tools/reply.rs (1)

476-490: Centralize text + cards merge policy to avoid drift.

This merge logic is now duplicated across multiple paths (src/tools/reply.rs, src/main.rs, src/messaging/webchat.rs, and cancelled-history extraction). Consider one shared helper on OutboundResponse for consistent behavior everywhere.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/tools/reply.rs` around lines 476 - 490, Create a single helper on
OutboundResponse (e.g., a method named flattened_text_or_merge or
merge_text_and_cards) that encapsulates the current logic: call
OutboundResponse::text_from_cards(cards) when self is RichMessage, then return
converted_content if card text is empty, card text if converted_content is
empty/whitespace, or the two joined with "\n\n" otherwise; use that method in
reply.rs (replace the inline block around OutboundResponse::RichMessage usage),
and replace the duplicated logic in main.rs, messaging/webchat.rs and the
cancelled-history extraction with calls to this new OutboundResponse helper so
the merge policy is centralized and consistent.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@src/agent/channel_history.rs`:
- Around line 172-193: The current extraction is gated on finding "content" as a
string; instead, always attempt to flatten cards from
tool_call.function.arguments.get("cards") via
serde_json::from_value::<Vec<crate::Card>>() and
crate::OutboundResponse::text_from_cards(&cards), then separately check for an
optional content string from
tool_call.function.arguments.get("content").as_str(); finally, return the
combined result: if both text and card_text are non-empty return
format!("{}\n\n{}", text, card_text), if only text return text, if only
card_text return card_text, otherwise None—update the logic around
tool_call.function.arguments, content, cards, and
crate::OutboundResponse::text_from_cards to implement this.

---

Nitpick comments:
In `@src/tools/reply.rs`:
- Around line 476-490: Create a single helper on OutboundResponse (e.g., a
method named flattened_text_or_merge or merge_text_and_cards) that encapsulates
the current logic: call OutboundResponse::text_from_cards(cards) when self is
RichMessage, then return converted_content if card text is empty, card text if
converted_content is empty/whitespace, or the two joined with "\n\n" otherwise;
use that method in reply.rs (replace the inline block around
OutboundResponse::RichMessage usage), and replace the duplicated logic in
main.rs, messaging/webchat.rs and the cancelled-history extraction with calls to
this new OutboundResponse helper so the merge policy is centralized and
consistent.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 7d7a2827-b5eb-4fbc-8e84-cce01808ca6b

📥 Commits

Reviewing files that changed from the base of the PR and between b7d5dd2 and dc3cf71.

📒 Files selected for processing (4)

• src/agent/channel_history.rs
• src/main.rs
• src/messaging/webchat.rs
• src/tools/reply.rs

[coderabbitai]

♻️ Duplicate comments (1)

src/agent/channel_history.rs (1)

172-194: ⚠️ Potential issue | 🟠 Major

Card extraction is still gated on content being a string.

On Line 172–174, the function still won’t preserve card-only reply payloads when content is absent or non-string. Parse cards independently, then merge whichever parts exist.

Suggested fix

-                    if let Some(content_value) = tool_call.function.arguments.get("content")
-                        && let Some(text) = content_value.as_str()
-                    {
-                        // Also extract card descriptions so the full response
-                        // (not just the short content text) is preserved in
-                        // conversation history for subsequent LLM turns.
-                        let cards = match tool_call.function.arguments.get("cards") {
-                            Some(v) => {
-                                serde_json::from_value::<Vec<crate::Card>>(v.clone())
-                                    .unwrap_or_else(|e| {
-                                        tracing::warn!(
-                                            error = %e,
-                                            "failed to deserialize cards from cancelled reply tool call; \
-                                             card content will be omitted from history"
-                                        );
-                                        Vec::new()
-                                    })
-                            }
-                            None => Vec::new(),
-                        };
-
-                        return Some(crate::OutboundResponse::text_with_cards(text, &cards));
-                    }
+                    let text = tool_call
+                        .function
+                        .arguments
+                        .get("content")
+                        .and_then(|value| value.as_str())
+                        .unwrap_or("");
+
+                    let cards = match tool_call.function.arguments.get("cards") {
+                        Some(value) => serde_json::from_value::<Vec<crate::Card>>(value.clone())
+                            .unwrap_or_else(|error| {
+                                tracing::warn!(
+                                    error = %error,
+                                    "failed to deserialize cards from cancelled reply tool call; \
+                                     card content will be omitted from history"
+                                );
+                                Vec::new()
+                            }),
+                        None => Vec::new(),
+                    };
+
+                    let merged = crate::OutboundResponse::text_with_cards(text, &cards);
+                    if !merged.trim().is_empty() {
+                        return Some(merged);
+                    }

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/agent/channel_history.rs` around lines 172 - 194, The current logic only
extracts cards when a string "content" exists, so card-only replies are dropped;
update the branch in the function handling tool_call.function.arguments to parse
cards unconditionally (use the existing
serde_json::from_value::<Vec<crate::Card>>(...) logic against
arguments.get("cards") regardless of "content"), then build the outbound
response by merging available parts: if content is a string use text, if cards
parsed non-empty include them, and return
crate::OutboundResponse::text_with_cards or the appropriate constructor with
just cards/text as needed (preserve the current warning behavior on deserialize
failure).

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Duplicate comments:
In `@src/agent/channel_history.rs`:
- Around line 172-194: The current logic only extracts cards when a string
"content" exists, so card-only replies are dropped; update the branch in the
function handling tool_call.function.arguments to parse cards unconditionally
(use the existing serde_json::from_value::<Vec<crate::Card>>(...) logic against
arguments.get("cards") regardless of "content"), then build the outbound
response by merging available parts: if content is a string use text, if cards
parsed non-empty include them, and return
crate::OutboundResponse::text_with_cards or the appropriate constructor with
just cards/text as needed (preserve the current warning behavior on deserialize
failure).

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 23e37f50-3b81-4c03-85ba-c82821109a52

📥 Commits

Reviewing files that changed from the base of the PR and between dc3cf71 and 3c9b6b6.

📒 Files selected for processing (5)

• src/agent/channel_history.rs
• src/lib.rs
• src/main.rs
• src/messaging/webchat.rs
• src/tools/reply.rs

🚧 Files skipped from review as they are similar to previous changes (2)

• src/messaging/webchat.rs
• src/main.rs