Jodinho commited on
Commit
62e6663
·
1 Parent(s): 178fd04

Latest changes

Browse files
.gitignore CHANGED
@@ -64,4 +64,5 @@ docs_research
64
  tools
65
  Real Estate Intelligence Layer.md
66
  grant_privileges.sql
67
- agent_prompt.md
 
 
64
  tools
65
  Real Estate Intelligence Layer.md
66
  grant_privileges.sql
67
+ agent_prompt.md
68
+ dynamic_tool_calling_features.txt
Frontend_API_Updates.md CHANGED
@@ -36,3 +36,11 @@ The LLM is now strictly instructed to output its replies using **Markdown** form
36
  - The frontend chat UI must be capable of rendering standard Markdown into HTML.
37
  - When the user asks to "download this data" or "export a CSV", the backend will generate the file, upload it to Appwrite, and return a Markdown link in the `reply` text (e.g., `[Download your CSV Report here](https://fra.cloud.appwrite.io/v1/storage/...)`).
38
  - The frontend's markdown parser should seamlessly render this as a standard clickable anchor tag (`<a>`) so the user can download the file.
 
 
 
 
 
 
 
 
 
36
  - The frontend chat UI must be capable of rendering standard Markdown into HTML.
37
  - When the user asks to "download this data" or "export a CSV", the backend will generate the file, upload it to Appwrite, and return a Markdown link in the `reply` text (e.g., `[Download your CSV Report here](https://fra.cloud.appwrite.io/v1/storage/...)`).
38
  - The frontend's markdown parser should seamlessly render this as a standard clickable anchor tag (`<a>`) so the user can download the file.
39
+
40
+ ## 4. Expanded Tool Registry & New Route
41
+
42
+ The backend has been upgraded with a new fallback route and additional Postgres RPC tools.
43
+
44
+ **Frontend Impact:**
45
+ - **PATH_C Fallback:** A new routing classification (`PATH_C`) is returned in `path_used` when the user asks a general real estate context question. The assistant will dynamically inject a disclaimer into the response text stating: *"Note: This is general market context, not live data from the Joule Dynamics tracking system."* No UI changes are strictly necessary, but be aware of the new `PATH_C` value in the response.
46
+ - **Robust Error Handling:** The assistant is explicitly instructed to never guess missing parameters or write ad-hoc queries. Missing parameters or invalid formats (such as UUID validation errors) will cleanly trigger the `suggested_actions` clarification flow to ask the user for correct details. Ensure the clarification buttons are rendered robustly.
services/groq_service.py CHANGED
@@ -10,27 +10,28 @@ session_history = {}
10
  groq_client = Groq(api_key=GROQ_API_KEY)
11
 
12
  ROUTER_PROMPT = """You are the classification router for the Joule Dynamics Real Estate Intelligence Layer.
13
- Analyze the user query and classify it into EXACTLY ONE of five classifications:
14
 
15
  1. "OUT_OF_SCOPE": Query asks about Leads, Lead-capture, Pricing Monitor data, general web crawling, or cross-system topics outside of the /real-estate page.
16
  2. "PATH_A": Query asks a live-data question (prices, spikes, availability, market averages, KPIs, specific listing rates).
17
  3. "PATH_B": Query asks a methodology/system design question (7-day average definition, 2-night check-in window, 4x daily scrape cadence, Vrbo status, World Cup strategy).
18
- 4. "BOTH": Query requires BOTH explaining a methodology concept AND fetching live data metrics.
19
- 5. "GREETING": User is saying hello, thanking the assistant, or making casual conversation without asking a specific question.
 
20
 
21
  Respond ONLY with valid JSON matching this schema:
22
- {"classification": "OUT_OF_SCOPE" | "PATH_A" | "PATH_B" | "BOTH" | "GREETING", "reason": "1-sentence justification"}
23
  """
24
 
25
  SYNTHESIS_PROMPT = """You are the B2B Real Estate Intelligence Assistant for Joule Dynamics.
26
  You provide precise data analysis to real estate investors and property managers reviewing short-term rental market performance.
27
 
28
  OPERATIONAL RULES:
29
- 1. NEVER FABRICATE DATA: Rely strictly on returned tool outputs or retrieved methodology chunks.
30
  2. ZERO GUESSING: If data or methodology is missing, state plainly: "I don't have that information in the current real estate scope."
31
  3. SCOPE BOUNDARY: If asked about Leads or Price Monitors, state that you are currently scoped exclusively to the Real Estate Rate Monitor.
32
- 4. FORMAT: Always format your final output in valid Markdown. Use bolding, lists, and tables to make data highly readable.
33
- 5. CLARIFICATION: If the user's request is ambiguous or if a tool is missing parameters, you can ask a clarifying question. To provide clickable options to the user, include a specific JSON block at the very end of your response exactly like this:
34
  ```json
35
  {"clarification_options": ["Option A", "Option B"]}
36
  ```
@@ -62,6 +63,8 @@ async def process_chat_message(user_query: str, session_id: str, session_context
62
  routing = {}
63
 
64
  classification = routing.get("classification", "PATH_A")
 
 
65
 
66
  # Guardrail: Immediate short-circuit if Out of Scope
67
  if classification == "OUT_OF_SCOPE":
@@ -94,6 +97,12 @@ async def process_chat_message(user_query: str, session_id: str, session_context
94
  messages = [
95
  {"role": "system", "content": SYNTHESIS_PROMPT}
96
  ]
 
 
 
 
 
 
97
  messages.extend(session_history[session_id])
98
 
99
  user_msg_content = f"User Context Filters: {json.dumps(session_context)}\nUser Query: {user_query}"
 
10
  groq_client = Groq(api_key=GROQ_API_KEY)
11
 
12
  ROUTER_PROMPT = """You are the classification router for the Joule Dynamics Real Estate Intelligence Layer.
13
+ Analyze the user query and classify it into EXACTLY ONE of six classifications:
14
 
15
  1. "OUT_OF_SCOPE": Query asks about Leads, Lead-capture, Pricing Monitor data, general web crawling, or cross-system topics outside of the /real-estate page.
16
  2. "PATH_A": Query asks a live-data question (prices, spikes, availability, market averages, KPIs, specific listing rates).
17
  3. "PATH_B": Query asks a methodology/system design question (7-day average definition, 2-night check-in window, 4x daily scrape cadence, Vrbo status, World Cup strategy).
18
+ 4. "PATH_C": Query asks a general real estate market context question that does not require live data metrics from our system.
19
+ 5. "BOTH": Query requires BOTH explaining a methodology concept AND fetching live data metrics.
20
+ 6. "GREETING": User is saying hello, thanking the assistant, or making casual conversation without asking a specific question.
21
 
22
  Respond ONLY with valid JSON matching this schema:
23
+ {"classification": "OUT_OF_SCOPE" | "PATH_A" | "PATH_B" | "PATH_C" | "BOTH" | "GREETING", "reason": "1-sentence justification"}
24
  """
25
 
26
  SYNTHESIS_PROMPT = """You are the B2B Real Estate Intelligence Assistant for Joule Dynamics.
27
  You provide precise data analysis to real estate investors and property managers reviewing short-term rental market performance.
28
 
29
  OPERATIONAL RULES:
30
+ 1. NEVER FABRICATE DATA: Rely strictly on returned tool outputs or retrieved methodology chunks. NEVER write ad-hoc SQL. You must exclusively use the registered tools provided.
31
  2. ZERO GUESSING: If data or methodology is missing, state plainly: "I don't have that information in the current real estate scope."
32
  3. SCOPE BOUNDARY: If asked about Leads or Price Monitors, state that you are currently scoped exclusively to the Real Estate Rate Monitor.
33
+ 4. FORMAT: Always format your final output in valid Markdown. Ensure you use tables, bold headers, and bulleted lists to make data highly readable.
34
+ 5. CLARIFICATION & ERRORS: If the user's request is ambiguous, a tool is missing parameters (like a market name or UUID), or if a tool returns an error message, DO NOT hallucinate inputs. Stop and provide a human-friendly response asking for clarification. To provide clickable options to the user, include a specific JSON block at the very end of your response exactly like this:
35
  ```json
36
  {"clarification_options": ["Option A", "Option B"]}
37
  ```
 
63
  routing = {}
64
 
65
  classification = routing.get("classification", "PATH_A")
66
+ if classification not in ["OUT_OF_SCOPE", "PATH_A", "PATH_B", "PATH_C", "BOTH", "GREETING"]:
67
+ classification = "PATH_A"
68
 
69
  # Guardrail: Immediate short-circuit if Out of Scope
70
  if classification == "OUT_OF_SCOPE":
 
97
  messages = [
98
  {"role": "system", "content": SYNTHESIS_PROMPT}
99
  ]
100
+ if classification == "PATH_C":
101
+ messages.append({
102
+ "role": "system",
103
+ "content": "Note: Because this is a PATH_C query, you must prepend a strict disclaimer stating: 'Note: This is general market context, not live data from the Joule Dynamics tracking system.'"
104
+ })
105
+
106
  messages.extend(session_history[session_id])
107
 
108
  user_msg_content = f"User Context Filters: {json.dumps(session_context)}\nUser Query: {user_query}"
services/supabase_service.py CHANGED
@@ -1,4 +1,5 @@
1
  import os
 
2
  from supabase import create_client, Client
3
  from services.embedding_service import get_embedding_model
4
  from config import SUPABASE_URL, SUPABASE_KEY
@@ -6,8 +7,26 @@ from config import SUPABASE_URL, SUPABASE_KEY
6
  supabase: Client = create_client(SUPABASE_URL, SUPABASE_KEY)
7
  embedder = get_embedding_model()
8
 
 
 
 
 
 
9
  async def execute_tool_rpc(func_name: str, args: dict) -> dict:
10
  """Executes a read-only Supabase RPC matching the tool schema."""
 
 
 
 
 
 
 
 
 
 
 
 
 
11
  try:
12
  res = supabase.rpc(func_name, args).execute()
13
  return {"status": "success", "data": res.data}
 
1
  import os
2
+ import re
3
  from supabase import create_client, Client
4
  from services.embedding_service import get_embedding_model
5
  from config import SUPABASE_URL, SUPABASE_KEY
 
7
  supabase: Client = create_client(SUPABASE_URL, SUPABASE_KEY)
8
  embedder = get_embedding_model()
9
 
10
+ def is_valid_uuid(val):
11
+ if not isinstance(val, str):
12
+ return False
13
+ return bool(re.match(r'^[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}$', val.lower()))
14
+
15
  async def execute_tool_rpc(func_name: str, args: dict) -> dict:
16
  """Executes a read-only Supabase RPC matching the tool schema."""
17
+ # Application layer clamping and validation
18
+ if func_name == "search_properties":
19
+ if "p_limit" in args and args["p_limit"] is not None:
20
+ args["p_limit"] = max(1, min(int(args["p_limit"]), 200))
21
+ elif func_name == "get_property_rate_changes":
22
+ if "days_param" in args and args["days_param"] is not None:
23
+ args["days_param"] = max(1, min(int(args["days_param"]), 90))
24
+ if "compare_window_days" in args and args["compare_window_days"] is not None:
25
+ args["compare_window_days"] = max(1, min(int(args["compare_window_days"]), 14))
26
+ elif func_name == "get_distance_km":
27
+ if not is_valid_uuid(args.get("property_a_id")) or not is_valid_uuid(args.get("property_b_id")):
28
+ return {"status": "error", "message": "Invalid UUID format. Both property IDs must be valid UUIDs."}
29
+
30
  try:
31
  res = supabase.rpc(func_name, args).execute()
32
  return {"status": "success", "data": res.data}
services/tools.py CHANGED
@@ -82,5 +82,69 @@ REAL_ESTATE_TOOLS = [
82
  "required": ["format", "content"]
83
  }
84
  }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
85
  }
86
  ]
 
82
  "required": ["format", "content"]
83
  }
84
  }
85
+ },
86
+ {
87
+ "type": "function",
88
+ "function": {
89
+ "name": "get_tracked_markets",
90
+ "description": "Fetch active markets being tracked, optionally filtered by platform.",
91
+ "parameters": {
92
+ "type": "object",
93
+ "properties": {
94
+ "p_platform": {"type": "string", "description": "Optional platform filter (e.g. 'Airbnb' or 'Vrbo')"}
95
+ },
96
+ "required": []
97
+ }
98
+ }
99
+ },
100
+ {
101
+ "type": "function",
102
+ "function": {
103
+ "name": "search_properties",
104
+ "description": "Search and filter properties by text, market, platform, bedrooms, or availability.",
105
+ "parameters": {
106
+ "type": "object",
107
+ "properties": {
108
+ "p_search": {"type": "string", "description": "Search text (name or uuid)"},
109
+ "p_market": {"type": "string", "description": "Market region name"},
110
+ "p_platform": {"type": "string", "description": "Platform (e.g. 'Airbnb')"},
111
+ "p_bedrooms": {"type": "integer", "description": "Number of bedrooms"},
112
+ "p_available": {"type": "boolean", "description": "Availability status"},
113
+ "p_limit": {"type": "integer", "default": 50, "description": "Number of results to return (max 200)"}
114
+ },
115
+ "required": []
116
+ }
117
+ }
118
+ },
119
+ {
120
+ "type": "function",
121
+ "function": {
122
+ "name": "get_property_rate_changes",
123
+ "description": "Show historical rate changes and trailing averages for a specific property over time.",
124
+ "parameters": {
125
+ "type": "object",
126
+ "properties": {
127
+ "property_search": {"type": "string", "description": "Property name or UUID (required)"},
128
+ "days_param": {"type": "integer", "default": 14, "description": "Lookback window in days (max 90)"},
129
+ "compare_window_days": {"type": "integer", "default": 1, "description": "Days to compare against for pct_change_vs_prev (max 14)"}
130
+ },
131
+ "required": ["property_search"]
132
+ }
133
+ }
134
+ },
135
+ {
136
+ "type": "function",
137
+ "function": {
138
+ "name": "get_distance_km",
139
+ "description": "Calculate the geographic distance in kilometers between two properties.",
140
+ "parameters": {
141
+ "type": "object",
142
+ "properties": {
143
+ "property_a_id": {"type": "string", "description": "UUID of the first property"},
144
+ "property_b_id": {"type": "string", "description": "UUID of the second property"}
145
+ },
146
+ "required": ["property_a_id", "property_b_id"]
147
+ }
148
+ }
149
  }
150
  ]