feat: Split button context by double vertical bar

Author: kieran-wilkinson-4

packages/cdk/prompts/systemPrompt.txt

@@ -1,59 +1,50 @@
-You are an AI assistant designed to provide expert guidance related to healthcare systems, data integration, and user setup. Leverage your contextual reasoning capabilities to synthesize complex information and provide evidence-based answers.
+You are an AI assistant designed to provide guidance and references from your knowledge base to help users make decisions when onboarding. It is *VERY* important you return *ALL* references, for user examination. 
 
+# Response
 ## Response Structure
-- *Summary*: 150 characters maximum, capturing core answer
-- *Answer* (use "mrkdown") (< 1k characters)
+- *Summary*: 100 characters maximum, capturing core answer
+- *Answer* (use "mrkdown") (< 500 characters)
 - Page break (use `------`)
 - \[Bibliography\]
 
-## Question Handling
-- Detect whether the query contains one or multiple questions
-- Split complex queries into individual sub-questions
-- Identify question type: factual, procedural, diagnostic, troubleshooting, or clarification-seeking
-- For multi-question queries: number sub-questions clearly (Q1, Q2, etc)
-
-## Analysis Workflow
-1. Break down question(s) into components; list explicit assumptions
-2. Identify information requirements and potential gaps
-3. Classify reference types needed:
-  - *Explicit*: direct quotes, named guidelines, official NHS/EPS documentation
-  - *Implicit*: inferred information (must caveat appropriately)
-4. Construct response using contextual reasoning:
-  - Connect findings logically across multiple documents
-  - Surface gaps, inconsistencies, or conflicting information
-  - Provide actionable steps with transparency
-
-## RAG & Knowledge Base Integration (ALWAYS QUERY FIRST)
-- Relevance threshold handling:
-  - Score ≥0.75 (High confidence): 
-    - Cite as: _"According to [Source Title]..."_
-  - Score 0.60-0.74 (Medium confidence): 
-    - Cite as: _"Based on available documentation (moderate confidence)..."_
-    - Add: "Recommend verification with latest [source type]"
-  - Score <0.60 (Low confidence): 
-    - Mark as inference: _"Documentation suggests... (low confidence)"_
-    - Add: "This interpretation requires verification"
-
 ## Formatting ("mrkdwn")
   a. *Bold* for:
     - Headings, subheadings: *Answer:*, *Bibliography:*
-    - Source names: *NHS Digital*, *EPS*
+    - Source names: *NHS England*, *EPS*
   b. _Italic_ for:
-    - Citations, references
-    - Document titles: _Integration Guide v3.2_
+    - Citations, references, document titles
   c. Block Quotes for:
     - Direct quotes >1 sentence
     - Technical specifications, parameters
-    - Example configurations
+    - Examples
   d. `Inline code` for:
     - System names, field names: `PrescriptionID`
     - Short technical terms: `HL7 FHIR`
   e. Links:
-    - Format: <https://example.com|Descriptive Name>
-    - Always test readability of link text
+    - Do not provide links
+
+# Thinking
+## Question Handling
+- Detect whether the query contains one or multiple questions
+- Split complex queries into individual sub-questions
+- Identify question type: factual, procedural, diagnostic, troubleshooting, or clarification-seeking
+- For multi-question queries: number sub-questions clearly (Q1, Q2, etc)
+
+## RAG & Knowledge Base Integration
+- Relevance threshold handling:
+  - Score > 0.80 (High confidence)
+  - Score 0.70 - 0.80 (Medium confidence)
+  - Score < 0.70 (Low confidence)
+
+## Corrections
+- Change _National Health Service Digital (NHSD)_ references to _National Health Service England (NHSE)_
+
+# Bibliography
+## Format*
+<cit>source number|summary title|link|filename|text snippet|reasoning</cit>\n
 
-# List items from <sr> at end of response as \[Bibliography\]
-  - *Format*: <cit>source number|summary title|link|filename|text snippet|</cit>\n
-  - Title should be less than 50 characters
-  - Review text snippets to check if they are useful
+## Requirements
+- Return **ALL** retrieved documents, their name and a text snippet, from "CONTEXT"
+- Snippet should be more than 50 characters
+- Title should be less than 50 characters
 

packages/cdk/prompts/userPrompt.txt

@@ -1,4 +1,6 @@
-<sr>$search_results$</sr>
-
-1. Using your knowledge around the National Health Service (NHS), Electronic Prescription Service (EPS) and the Fast Healthcare Interoperability Resources' (FHIR) onboarding, Supplier Conformance Assessment List (SCAL), NHS SPINE, APIs, developer guides and error resolution; please answer the following question:
+# QUERY
 {{user_query}}
+
+# CONTEXT
+## Results $search_results$
+## LIST ALL RESULTS IN TABLE

packages/slackBotFunction/app/slack/slack_events.py

@@ -165,15 +165,19 @@ def _create_feedback_blocks(
     if citations is None or len(citations) == 0:
         logger.info("No citations")
     else:
+        invalid_body = "No document excerpt available."
         for i, citation in enumerate(citations):
             logger.info("Creating citation", extra={"Citation": citation})
-            # Create citation blocks
-            # keys = ["sourceNumber", "title", "link", "filename", "reference_text"]
+            # Create citation blocks ["sourceNumber", "title", "link", "filename", "reference_text"]
             title = citation.get("title") or citation.get("filename") or "Source"
-            body = citation.get("reference_text") or "No citation text available."
+            body = citation.get("reference_text") or invalid_body
             citation_link = citation.get("link") or ""
             source_number = (citation.get("source_number", "0")).replace("\n", "")
 
+            # If snippet is from dev table or is a single word, skip
+            if re.fullmatch(r"[A-Za-z0-9-_]+", body.strip()) >= 0:
+                body = invalid_body
+
             # Buttons can only be 75 characters long, truncate to be safe
             button_text = f"[{source_number}] {title}"
             button = {
@@ -436,7 +440,7 @@ def process_slack_message(event: Dict[str, Any], event_id: str, client: WebClien
             raw_citations = re.compile(r"<cit\b[^>]*>(.*?)</cit>", re.DOTALL | re.IGNORECASE).findall(citation_block)
             if len(raw_citations) > 0:
                 logger.info("Found citation(s)", extra={"Raw Citations": raw_citations})
-                citations = [dict(zip(prompt_value_keys, citation.split("|"))) for citation in raw_citations]
+                citations = [dict(zip(prompt_value_keys, citation.split("||"))) for citation in raw_citations]
         logger.info("Parsed citation(s)", extra={"citations": citations})
 
         # Post the answer (plain) to get message_ts
@@ -619,20 +623,22 @@ def open_citation(channel: str, timestamp: str, message: Any, params: Dict[str,
     logger.info("Opening citation", extra={"channel": channel, "timestamp": timestamp})
     try:
         # Citation details
-        title: str = params.get("title", "No title available.")
-        body: str = params.get("body", "No citation text available.")
+        title: str = params.get("title", "No title available.").strip()
+        body: str = params.get("body", "No citation text available.").strip()
         source_number: str = params.get("source_number")
-        link: str = params.get("link", "")
 
         # Remove any existing citation block/divider
         blocks = message.get("blocks", [])
         blocks = [b for b in blocks if b.get("block_id") not in ["citation_block", "citation_divider"]]
 
         # Format text
         title = f"*{title.replace('\n', '')}*"
-        body = f"> {body.replace('\n', '\n> ')}"
+        if body and len(body) > 0:
+            body = f"> {body.replace('\n', '\n> ')}"  # Block quote
+            body = re.sub(r"\[([^\]]+)\]\(([^\)]+)\)", r"<\1|\2>", body)  # Convert links
+            body = body.replace("»", "")  # Remove double chevrons
 
-        current_id = f"cite_{source_number}"
+        current_id = f"cite_{source_number}".strip()
         selected = False
 
         # Reset all button styles, then set the clicked one
@@ -657,14 +663,7 @@ def open_citation(channel: str, timestamp: str, message: Any, params: Dict[str,
         if selected:
             citation_block = {
                 "type": "section",
-                "text": {
-                    "type": "mrkdwn",
-                    "text": (
-                        f"{title}\n\n{body}\n\n<{link}|View Source>"
-                        if link and link != "none"
-                        else f"{title}\n\n{body}"
-                    ),
-                },
+                "text": {"type": "mrkdwn", "text": f"{title}\n\n{body}"},
                 "block_id": "citation_block",
             }
             feedback_index = next(