feat: extract text content from message blocks in conversations_history

[derodero24]

Summary

Extract text from Slack Block Kit structures in messages retrieved via conversations_history and conversations_replies, so that block-only messages (emails, bot notifications, app messages) are no longer returned empty.

Closes #186

Problem

Messages using Block Kit format (e.g., Slack Email integration, Grafana/Datadog alerts) have an empty Text field. The existing AttachmentsTo2CSV extracts text from attachment fields but ignores:

• Top-level blocks array
• attachments[].blocks array

Solution

1. BlocksToText function in pkg/text/text_processor.go extracts text from common block types:
   • header.text
   • section.text and section.fields
   • rich_text (sections, lists, quotes, preformatted)
   • context.elements[].text
2. Top-level blocks: When msg.Text is empty, fall back to BlocksToText(msg.Blocks) — avoids duplication since Slack typically populates text as a plaintext fallback of blocks
3. Attachment blocks: AttachmentToText now also extracts att.Blocks

Changes

• pkg/text/text_processor.go: Add BlocksToText and rich text helper functions; call BlocksToText in AttachmentToText
• pkg/handler/conversations.go: Use block text as fallback in convertMessagesFromHistory and convertMessagesFromSearch
• pkg/text/text_processor_test.go: Add 15 test cases covering all supported block types, edge cases, and attachment integration

Testing

• All existing unit tests pass (go test -run TestUnit ./...)
• New tests cover: empty blocks, header, section (with text/fields/nil text), context, rich_text (text/link/broadcast), multiple blocks combined, non-text blocks skipped, attachment with blocks only, attachment with both text and blocks

[korotovsky]

Could you please provide a couple of screenshots/examples that illustrate before/after results?

[derodero24]

@korotovsky Sure! Here are before/after examples based on real Slack messages.

Example 1: CloudWatch Alarm (attachment with Block Kit only)

This is a real message from an AWS monitoring bot. The message has text: "" and attachments[0] contains only Block Kit blocks (no legacy title/text fields).

Attachment JSON (simplified):

{
  "text": "",
  "attachments": [{
    "title": "",
    "text": "",
    "blocks": [
      {"type": "section", "text": {"type": "mrkdwn", "text": "*<URL|:rotating_light: CloudWatch Alarm | MyAlarm | ap-northeast-1>*"}},
      {"type": "section", "text": {"type": "mrkdwn", "text": "Threshold Crossed: 1 out of the last 1 datapoints [1.0] was greater than or equal to the threshold (1.0)"}},
      {"type": "actions", "elements": [...]},
      {"type": "section", "fields": [{"type": "mrkdwn", "text": "*Namespace*\nMyNamespace"}, {"type": "mrkdwn", "text": "*Metric*\nMyMetric"}]},
      {"type": "section", "fields": [{"type": "mrkdwn", "text": "*Alarm State*\nALARM"}]},
      {"type": "image", "...": "..."},
      {"type": "context", "elements": [{"type": "mrkdwn", "text": "<!date^1738900000^{date_short_pretty} at {time}|2025-02-07>"}]}
    ]
  }]
}

Before — conversations_history CSV text column:

(empty)

After — Block Kit content is extracted from attachments[].blocks:

Blocks: *URL - :rotating_light: CloudWatch Alarm | MyAlarm | ap-northeast-1* Threshold Crossed: 1 out of the last 1 datapoints [1.0] was greater than or equal to the threshold [1.0] *Namespace* MyNamespace *Metric* MyMetric *Alarm State* ALARM <!date^1738900000^{date_short_pretty} at {time}|2025-02-07>

Example 2: Datadog Alert (legacy attachment format — no change)

A Datadog alert uses legacy attachment fields (title + text) without Block Kit blocks. Behavior is unchanged:

{
  "text": "",
  "attachments": [{
    "title": "Triggered: ServerError - /ecs/my-service",
    "text": "Host: /ecs/my-service\nLog status: error\nMore than 1 log event matched..."
  }]
}

Before and After — Same output (already extracted via legacy fields):

Title: Triggered: ServerError - /ecs/my-service; Text: Host: /ecs/my-service Log status: error More than 1 log event matched...

Example 3: Top-level Block Kit message (bot with text: "")

Some bots send messages with text: "" and content only in top-level blocks:

{
  "text": "",
  "blocks": [
    {"type": "header", "text": {"type": "plain_text", "text": "Deploy Complete"}},
    {"type": "section", "text": {"type": "mrkdwn", "text": "Service *my-api* deployed to production"}},
    {"type": "context", "elements": [{"type": "mrkdwn", "text": "Deployed by CI/CD pipeline"}]}
  ]
}

Before — text column:

(empty)

After — Block Kit text extracted as fallback (only when msg.Text is empty):

Deploy Complete Service *my-api* deployed to production Deployed by CI/CD pipeline

Design notes

• Top-level msg.Blocks: Extracted only when msg.Text is empty (Slack usually provides a plaintext fallback in text, so this avoids duplication)
• attachments[].Blocks: Always extracted as supplementary info (prefixed with Blocks:)
• Non-text blocks (divider, image, actions) are skipped

[korotovsky]

@derodero24 thank you, and how does serialized CSV message look like in such cases?

[derodero24]

@korotovsky Here's the actual serialized CSV output generated by running the code against representative Block Kit messages:

Example 1: CloudWatch Alarm (attachment with blocks only)

Before:

MsgID,UserID,UserName,RealName,Channel,ThreadTs,Text,Time,BotName,Cursor
1770433087.532289,U001,aws-bot,Amazon Q Developer,C001,,,2026-02-07T02:58:02Z,Amazon Q Developer,

After:

MsgID,UserID,UserName,RealName,Channel,ThreadTs,Text,Time,BotName,Cursor
1770433087.532289,U001,aws-bot,Amazon Q Developer,C001,,"Blocks: https://console.aws.amazon.com/cloudwatch - :rotating_light: CloudWatch Alarm MyAlarm ap-northeast-1 Account: 123456789012, Threshold Crossed: 1 out of the last 1 datapoints 1.0 07/02/26 02:57:00 was greater than or equal to the threshold 1.0 Namespace MyNamespace Metric MyMetric Alarm State ALARM",2026-02-07T02:58:02Z,Amazon Q Developer,

Example 2: Bot message with top-level blocks

Before:

MsgID,UserID,UserName,RealName,Channel,ThreadTs,Text,Time,BotName,Cursor
1770400000.000001,U002,deploy-bot,Deploy Bot,C002,,,2026-02-07T10:00:00Z,Deploy Bot,

After:

MsgID,UserID,UserName,RealName,Channel,ThreadTs,Text,Time,BotName,Cursor
1770400000.000001,U002,deploy-bot,Deploy Bot,C002,,Deploy Complete Service my-api deployed to production Deployed by CI/CD pipeline,2026-02-07T10:00:00Z,Deploy Bot,

Example 3: Legacy attachment (no blocks) — unchanged

MsgID,UserID,UserName,RealName,Channel,ThreadTs,Text,Time,BotName,Cursor
1770433155.658769,U003,datadog,Datadog,C001,,Title: Triggered: ServerError Text: Host: /ecs/my-service Log status: error More than 1 log event matched,2026-02-07T02:59:15Z,Datadog,