Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -14,6 +14,8 @@ from gtts import gTTS
|
|
| 14 |
DATA_FILE = "legislation_rules.json"
|
| 15 |
RULINGS_FILE = "rulings_log.json"
|
| 16 |
|
|
|
|
|
|
|
| 17 |
def load_data(file_path):
|
| 18 |
if os.path.exists(file_path):
|
| 19 |
with open(file_path, 'r') as f:
|
|
@@ -34,7 +36,6 @@ def extract_text_from_any(file_obj):
|
|
| 34 |
if file_obj is None: return ""
|
| 35 |
file_path = file_obj.name if hasattr(file_obj, 'name') else file_obj
|
| 36 |
if not os.path.exists(file_path): return ""
|
| 37 |
-
|
| 38 |
ext = os.path.splitext(file_path)[1].lower()
|
| 39 |
text = ""
|
| 40 |
try:
|
|
@@ -74,12 +75,10 @@ def process_rule_document(file_obj, act_name):
|
|
| 74 |
return add_rule_manually(act_name, os.path.basename(file_obj.name), text, f"File: {os.path.basename(file_obj.name)}")
|
| 75 |
|
| 76 |
def generate_tts(text):
|
| 77 |
-
|
| 78 |
-
if not text or len(text) < 5:
|
| 79 |
-
return None
|
| 80 |
try:
|
| 81 |
clean_text = text.replace("#", "").replace("*", "").replace("`", "")
|
| 82 |
-
tts = gTTS(text=clean_text[:1500], lang='en')
|
| 83 |
filename = f"ruling_audio_{int(datetime.now().timestamp())}.mp3"
|
| 84 |
tts.save(filename)
|
| 85 |
return filename
|
|
@@ -90,31 +89,11 @@ def generate_tts(text):
|
|
| 90 |
def rule_on_issue(issue_description, issue_file, llm_endpoint, llm_key, llm_model):
|
| 91 |
file_text = extract_text_from_any(issue_file)
|
| 92 |
combined_issue = f"{issue_description}\n\n[EXTRACTED FROM DOCUMENT]:\n{file_text}".strip()
|
| 93 |
-
|
| 94 |
-
if not combined_issue:
|
| 95 |
-
return "Error: No issue description or document provided.", "", None
|
| 96 |
-
|
| 97 |
rules = load_data(DATA_FILE)
|
| 98 |
-
if not rules:
|
| 99 |
-
return "Error: Dataset is empty. Add rules (CRA, FCA, etc.) first.", "", None
|
| 100 |
-
|
| 101 |
context = "\n".join([f"[{r['act']} - {r['section_title']}]: {r['text'][:600]}..." for r in rules[:20]])
|
| 102 |
-
|
| 103 |
-
prompt = f"""
|
| 104 |
-
You are a Regulatory Ruling Engine.
|
| 105 |
-
ISSUE TO ANALYZE:
|
| 106 |
-
{combined_issue}
|
| 107 |
-
|
| 108 |
-
DETERMINISTIC REFERENCE RULES:
|
| 109 |
-
{context}
|
| 110 |
-
|
| 111 |
-
TASK:
|
| 112 |
-
1. Identify applicable rules.
|
| 113 |
-
2. Provide a 'Formal Ruling' (Compliant/Non-Compliant/Warning).
|
| 114 |
-
3. Cite Rule Titles and Deterministic IDs.
|
| 115 |
-
4. Provide forensic justification.
|
| 116 |
-
"""
|
| 117 |
-
|
| 118 |
ruling_text = "LLM Inference required for automated ruling."
|
| 119 |
if llm_endpoint and llm_key:
|
| 120 |
try:
|
|
@@ -122,23 +101,12 @@ def rule_on_issue(issue_description, issue_file, llm_endpoint, llm_key, llm_mode
|
|
| 122 |
payload = {"model": llm_model, "messages": [{"role": "user", "content": prompt}], "temperature": 0}
|
| 123 |
response = requests.post(f"{llm_endpoint.rstrip('/')}/v1/chat/completions", json=payload, headers=headers, timeout=45)
|
| 124 |
ruling_text = response.json()['choices'][0]['message']['content']
|
| 125 |
-
except Exception as e:
|
| 126 |
-
ruling_text = f"Inference Error: {str(e)}"
|
| 127 |
-
|
| 128 |
-
# Generate Audio for the Ruling
|
| 129 |
audio_file = generate_tts(ruling_text)
|
| 130 |
-
|
| 131 |
-
# Log the ruling
|
| 132 |
rulings_log = load_data(RULINGS_FILE)
|
| 133 |
-
new_ruling = {
|
| 134 |
-
"timestamp": datetime.now().isoformat(),
|
| 135 |
-
"issue_summary": combined_issue[:200],
|
| 136 |
-
"ruling": ruling_text,
|
| 137 |
-
"ruling_hash": hashlib.sha256(ruling_text.encode()).hexdigest()
|
| 138 |
-
}
|
| 139 |
rulings_log.append(new_ruling)
|
| 140 |
save_data(rulings_log, RULINGS_FILE)
|
| 141 |
-
|
| 142 |
return ruling_text, f"Ruling Logged (ID: {new_ruling['ruling_hash'][:8]})", audio_file
|
| 143 |
|
| 144 |
def sync_all(token, dataset_id):
|
|
@@ -158,6 +126,25 @@ def view_stats():
|
|
| 158 |
for r in data: acts[r['act']] = acts.get(r['act'], 0) + 1
|
| 159 |
return "\n".join([f"{k}: {v}" for k,v in acts.items()])
|
| 160 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 161 |
with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
| 162 |
gr.Markdown("# โ๏ธ Legislation & FCA Ruling Manager")
|
| 163 |
|
|
@@ -193,10 +180,18 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
|
| 193 |
stats_out = gr.Textbox(label="Dataset Inventory", value=view_stats())
|
| 194 |
refresh_btn = gr.Button("Refresh Inventory")
|
| 195 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 196 |
with gr.Tab("โ๏ธ Sync & Audit"):
|
| 197 |
hf_t = gr.Textbox(label="HF Token", type="password"); hf_d = gr.Textbox(label="Dataset ID")
|
| 198 |
s_btn = gr.Button("Sync Everything to Dataset")
|
| 199 |
s_out = gr.Textbox(label="Sync Status")
|
|
|
|
|
|
|
|
|
|
| 200 |
|
| 201 |
rule_btn.click(fn=rule_on_issue, inputs=[issue_desc, issue_file, end, key, mod], outputs=[ruling_out, log_status, audio_out])
|
| 202 |
doc_btn.click(fn=process_rule_document, inputs=[doc_in, doc_act], outputs=op_status)
|
|
|
|
| 14 |
DATA_FILE = "legislation_rules.json"
|
| 15 |
RULINGS_FILE = "rulings_log.json"
|
| 16 |
|
| 17 |
+
# --- CORE LOGIC ---
|
| 18 |
+
|
| 19 |
def load_data(file_path):
|
| 20 |
if os.path.exists(file_path):
|
| 21 |
with open(file_path, 'r') as f:
|
|
|
|
| 36 |
if file_obj is None: return ""
|
| 37 |
file_path = file_obj.name if hasattr(file_obj, 'name') else file_obj
|
| 38 |
if not os.path.exists(file_path): return ""
|
|
|
|
| 39 |
ext = os.path.splitext(file_path)[1].lower()
|
| 40 |
text = ""
|
| 41 |
try:
|
|
|
|
| 75 |
return add_rule_manually(act_name, os.path.basename(file_obj.name), text, f"File: {os.path.basename(file_obj.name)}")
|
| 76 |
|
| 77 |
def generate_tts(text):
|
| 78 |
+
if not text or len(text) < 5: return None
|
|
|
|
|
|
|
| 79 |
try:
|
| 80 |
clean_text = text.replace("#", "").replace("*", "").replace("`", "")
|
| 81 |
+
tts = gTTS(text=clean_text[:1500], lang='en')
|
| 82 |
filename = f"ruling_audio_{int(datetime.now().timestamp())}.mp3"
|
| 83 |
tts.save(filename)
|
| 84 |
return filename
|
|
|
|
| 89 |
def rule_on_issue(issue_description, issue_file, llm_endpoint, llm_key, llm_model):
|
| 90 |
file_text = extract_text_from_any(issue_file)
|
| 91 |
combined_issue = f"{issue_description}\n\n[EXTRACTED FROM DOCUMENT]:\n{file_text}".strip()
|
| 92 |
+
if not combined_issue: return "Error: No issue description or document provided.", "", None
|
|
|
|
|
|
|
|
|
|
| 93 |
rules = load_data(DATA_FILE)
|
| 94 |
+
if not rules: return "Error: Dataset is empty. Add rules (CRA, FCA, etc.) first.", "", None
|
|
|
|
|
|
|
| 95 |
context = "\n".join([f"[{r['act']} - {r['section_title']}]: {r['text'][:600]}..." for r in rules[:20]])
|
| 96 |
+
prompt = f"You are a Regulatory Ruling Engine.\nISSUE TO ANALYZE:\n{combined_issue}\nDETERMINISTIC REFERENCE RULES:\n{context}\nTASK:\n1. Identify applicable rules.\n2. Provide a 'Formal Ruling' (Compliant/Non-Compliant/Warning).\n3. Cite Rule Titles and Deterministic IDs.\n4. Provide forensic justification."
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 97 |
ruling_text = "LLM Inference required for automated ruling."
|
| 98 |
if llm_endpoint and llm_key:
|
| 99 |
try:
|
|
|
|
| 101 |
payload = {"model": llm_model, "messages": [{"role": "user", "content": prompt}], "temperature": 0}
|
| 102 |
response = requests.post(f"{llm_endpoint.rstrip('/')}/v1/chat/completions", json=payload, headers=headers, timeout=45)
|
| 103 |
ruling_text = response.json()['choices'][0]['message']['content']
|
| 104 |
+
except Exception as e: ruling_text = f"Inference Error: {str(e)}"
|
|
|
|
|
|
|
|
|
|
| 105 |
audio_file = generate_tts(ruling_text)
|
|
|
|
|
|
|
| 106 |
rulings_log = load_data(RULINGS_FILE)
|
| 107 |
+
new_ruling = {"timestamp": datetime.now().isoformat(), "issue_summary": combined_issue[:200], "ruling": ruling_text, "ruling_hash": hashlib.sha256(ruling_text.encode()).hexdigest()}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 108 |
rulings_log.append(new_ruling)
|
| 109 |
save_data(rulings_log, RULINGS_FILE)
|
|
|
|
| 110 |
return ruling_text, f"Ruling Logged (ID: {new_ruling['ruling_hash'][:8]})", audio_file
|
| 111 |
|
| 112 |
def sync_all(token, dataset_id):
|
|
|
|
| 126 |
for r in data: acts[r['act']] = acts.get(r['act'], 0) + 1
|
| 127 |
return "\n".join([f"{k}: {v}" for k,v in acts.items()])
|
| 128 |
|
| 129 |
+
# --- GRADIO UI ---
|
| 130 |
+
|
| 131 |
+
USER_MANUAL_MD = """
|
| 132 |
+
# โ๏ธ Legislation & FCA Ruling Manager: User Manual
|
| 133 |
+
|
| 134 |
+
## ๐ ๏ธ Building Your Rulebase
|
| 135 |
+
1. **Manual Entry**: Go to **Manage Rules** to paste specific sections from acts.
|
| 136 |
+
2. **Document Ingestion**: Upload **PDF, DOCX, or TXT** files of internal policies or acts.
|
| 137 |
+
3. **FCA Automation**: Use the **FCA Guidelines** tab to pull the **12 Principles for Businesses (PRIN)**.
|
| 138 |
+
|
| 139 |
+
## ๐๏ธ Ruling on Issues
|
| 140 |
+
1. **Describing the Issue**: Go to **Rule on Issues**. Upload an **Issue Document** or type a description.
|
| 141 |
+
2. **Generating the Ruling**: Ensure your **Inference Settings** are configured, then click **Generate Formal Ruling**.
|
| 142 |
+
3. **Audible Delivery (TTS)**: Click play on the **Audio Player** to hear the ruling read aloud.
|
| 143 |
+
|
| 144 |
+
## โ๏ธ Sync & Audit
|
| 145 |
+
- Go to **Sync & Audit**, enter your **HF Write Token** and **Dataset ID**, and click **Sync Everything**.
|
| 146 |
+
"""
|
| 147 |
+
|
| 148 |
with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
| 149 |
gr.Markdown("# โ๏ธ Legislation & FCA Ruling Manager")
|
| 150 |
|
|
|
|
| 180 |
stats_out = gr.Textbox(label="Dataset Inventory", value=view_stats())
|
| 181 |
refresh_btn = gr.Button("Refresh Inventory")
|
| 182 |
|
| 183 |
+
with gr.Tab("๐ฆ FCA Guidelines"):
|
| 184 |
+
gr.Markdown("### ๐ ๏ธ FCA Handbook Automation")
|
| 185 |
+
fca_btn = gr.Button("Ingest FCA PRIN Principles", variant="secondary")
|
| 186 |
+
fca_status = gr.Textbox(label="FCA Ingestion Status")
|
| 187 |
+
|
| 188 |
with gr.Tab("โ๏ธ Sync & Audit"):
|
| 189 |
hf_t = gr.Textbox(label="HF Token", type="password"); hf_d = gr.Textbox(label="Dataset ID")
|
| 190 |
s_btn = gr.Button("Sync Everything to Dataset")
|
| 191 |
s_out = gr.Textbox(label="Sync Status")
|
| 192 |
+
|
| 193 |
+
with gr.Tab("๐ User Manual"):
|
| 194 |
+
gr.Markdown(USER_MANUAL_MD)
|
| 195 |
|
| 196 |
rule_btn.click(fn=rule_on_issue, inputs=[issue_desc, issue_file, end, key, mod], outputs=[ruling_out, log_status, audio_out])
|
| 197 |
doc_btn.click(fn=process_rule_document, inputs=[doc_in, doc_act], outputs=op_status)
|