Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,3 +1,4 @@
|
|
|
|
1 |
# === Imports ===
|
2 |
import os
|
3 |
import re
|
@@ -17,16 +18,7 @@ client = openai.OpenAI(api_key=openai_api_key)
|
|
17 |
openrouter_key = os.environ.get("OPENROUTER")
|
18 |
openrouter = openai.OpenAI(api_key=openrouter_key, base_url="https://openrouter.ai/api/v1")
|
19 |
|
20 |
-
# --- OCI Object Storage
|
21 |
-
#oci_config = {
|
22 |
-
# "user": os.environ.get("OCI_USER"),
|
23 |
-
# "tenancy": os.environ.get("OCI_TENANCY"),
|
24 |
-
# "fingerprint": os.environ.get("OCI_FINGERPRINT"),
|
25 |
-
# "region": os.environ.get("OCI_REGION"),
|
26 |
-
# "key_content": os.environ.get("OCI_PRIVATE_KEY"),
|
27 |
-
#}
|
28 |
-
|
29 |
-
# === OCI Object Storage Setup ===
|
30 |
oci_config = {
|
31 |
"user": os.environ.get("OCI_USER"),
|
32 |
"tenancy": os.environ.get("OCI_TENANCY"),
|
@@ -43,11 +35,6 @@ try:
|
|
43 |
except Exception as e:
|
44 |
print("Failed to initialize OCI Object Storage client:", e)
|
45 |
|
46 |
-
|
47 |
-
namespace = os.environ.get("OCI_NAMESPACE")
|
48 |
-
bucket_name = os.environ.get("OCI_BUCKET_NAME")
|
49 |
-
object_storage = oci.object_storage.ObjectStorageClient(oci_config)
|
50 |
-
|
51 |
# --- Exadata Specs ---
|
52 |
exadata_specs = {
|
53 |
"X7": {"Quarter Rack": {"max_iops": 350000, "max_throughput": 25}, "Half Rack": {"max_iops": 700000, "max_throughput": 50}, "Full Rack": {"max_iops": 1400000, "max_throughput": 100}},
|
@@ -75,7 +62,7 @@ def upload_awr_file(file_obj):
|
|
75 |
filename = os.path.basename(file_obj.name)
|
76 |
content = file_obj.read()
|
77 |
object_storage.put_object(namespace, bucket_name, filename, content)
|
78 |
-
return f"
|
79 |
|
80 |
def list_awr_files():
|
81 |
try:
|
@@ -99,27 +86,145 @@ def compare_awrs(file_list, llm_model):
|
|
99 |
for fname in file_list:
|
100 |
content = get_awr_file_text(fname)
|
101 |
combined_text += f"\n=== AWR: {fname} ===\n{content[:3000]}...\n"
|
102 |
-
prompt = f"""
|
103 |
-
You are a senior Oracle performance engineer. You will compare multiple AWR reports and highlight:
|
104 |
- Key differences in workload or system behavior
|
105 |
- Major trends or anomalies
|
106 |
- Which report shows better performance and why
|
107 |
- Exadata-specific metrics like Smart Scan, Flash I/O
|
108 |
- Suggestions to unify or improve system behavior
|
|
|
109 |
AWR Reports:
|
110 |
{combined_text}
|
111 |
"""
|
112 |
response = client.chat.completions.create(
|
113 |
model=llm_model,
|
114 |
-
messages=[
|
115 |
-
|
|
|
|
|
116 |
)
|
117 |
return response.choices[0].message.content.strip()
|
118 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
119 |
# === Gradio UI ===
|
120 |
with gr.Blocks() as demo:
|
121 |
with gr.Tab("Manual AWR Analysis"):
|
122 |
-
gr.Markdown("#
|
123 |
awr_text = gr.Textbox(label="Paste AWR Report", lines=30)
|
124 |
threshold = gr.Slider(0, 5, value=3, step=1, label="Correctness Threshold (Stars)")
|
125 |
performance_test_mode = gr.Checkbox(label="Performance Test Mode")
|
@@ -137,7 +242,6 @@ with gr.Blocks() as demo:
|
|
137 |
rating = gr.Textbox(label="Rater", lines=3)
|
138 |
retry_status = gr.Textbox(label="Retry Status")
|
139 |
|
140 |
-
from your_existing_code import process_awr # Replace with actual import or include function here
|
141 |
analyze_btn.click(process_awr,
|
142 |
inputs=[awr_text, threshold, performance_test_mode, exadata_model, rack_size, llm_selector],
|
143 |
outputs=[output, health, rating, retry_status])
|
@@ -147,7 +251,7 @@ with gr.Blocks() as demo:
|
|
147 |
upload_status = gr.Textbox(label="Upload Status")
|
148 |
upload_file.upload(fn=upload_awr_file, inputs=upload_file, outputs=upload_status)
|
149 |
|
150 |
-
refresh_button = gr.Button("
|
151 |
file_multiselect = gr.Dropdown(choices=[], label="Select AWR Files", multiselect=True)
|
152 |
refresh_button.click(fn=lambda: gr.update(choices=list_awr_files()), outputs=file_multiselect)
|
153 |
|
@@ -157,4 +261,5 @@ with gr.Blocks() as demo:
|
|
157 |
|
158 |
if __name__ == "__main__":
|
159 |
demo.launch(debug=True)
|
160 |
-
|
|
|
|
1 |
+
|
2 |
# === Imports ===
|
3 |
import os
|
4 |
import re
|
|
|
18 |
openrouter_key = os.environ.get("OPENROUTER")
|
19 |
openrouter = openai.OpenAI(api_key=openrouter_key, base_url="https://openrouter.ai/api/v1")
|
20 |
|
21 |
+
# --- OCI Object Storage Setup ---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
22 |
oci_config = {
|
23 |
"user": os.environ.get("OCI_USER"),
|
24 |
"tenancy": os.environ.get("OCI_TENANCY"),
|
|
|
35 |
except Exception as e:
|
36 |
print("Failed to initialize OCI Object Storage client:", e)
|
37 |
|
|
|
|
|
|
|
|
|
|
|
38 |
# --- Exadata Specs ---
|
39 |
exadata_specs = {
|
40 |
"X7": {"Quarter Rack": {"max_iops": 350000, "max_throughput": 25}, "Half Rack": {"max_iops": 700000, "max_throughput": 50}, "Full Rack": {"max_iops": 1400000, "max_throughput": 100}},
|
|
|
62 |
filename = os.path.basename(file_obj.name)
|
63 |
content = file_obj.read()
|
64 |
object_storage.put_object(namespace, bucket_name, filename, content)
|
65 |
+
return f"β
Uploaded {filename}"
|
66 |
|
67 |
def list_awr_files():
|
68 |
try:
|
|
|
86 |
for fname in file_list:
|
87 |
content = get_awr_file_text(fname)
|
88 |
combined_text += f"\n=== AWR: {fname} ===\n{content[:3000]}...\n"
|
89 |
+
prompt = f"""You are a senior Oracle performance engineer. You will compare multiple AWR reports and highlight:
|
|
|
90 |
- Key differences in workload or system behavior
|
91 |
- Major trends or anomalies
|
92 |
- Which report shows better performance and why
|
93 |
- Exadata-specific metrics like Smart Scan, Flash I/O
|
94 |
- Suggestions to unify or improve system behavior
|
95 |
+
|
96 |
AWR Reports:
|
97 |
{combined_text}
|
98 |
"""
|
99 |
response = client.chat.completions.create(
|
100 |
model=llm_model,
|
101 |
+
messages=[
|
102 |
+
{"role": "system", "content": "You are a comparative AWR analysis expert."},
|
103 |
+
{"role": "user", "content": prompt}
|
104 |
+
]
|
105 |
)
|
106 |
return response.choices[0].message.content.strip()
|
107 |
|
108 |
+
# === AGENTS ===
|
109 |
+
class CriticalAnalyzerAgent:
|
110 |
+
def analyze(self, content, performance_test_mode, exadata_model, rack_size, llm_model):
|
111 |
+
cleaned_content = clean_awr_content(content)
|
112 |
+
if len(cleaned_content) > 128000:
|
113 |
+
cleaned_content = cleaned_content[:128000] + "\n\n[TRUNCATED]..."
|
114 |
+
|
115 |
+
prompt = f"""You are an expert Oracle DBA performance analyst specialized in AWR + Exadata.
|
116 |
+
|
117 |
+
Please perform advanced analysis on the following report:
|
118 |
+
|
119 |
+
======== AWR REPORT START ========
|
120 |
+
{cleaned_content}
|
121 |
+
======== AWR REPORT END ========
|
122 |
+
|
123 |
+
Required Output:
|
124 |
+
- Performance Summary (with metric values)
|
125 |
+
- Detailed Bottlenecks + Risks (quantified)
|
126 |
+
- Forecast + Predictions
|
127 |
+
- Monitoring Recommendations
|
128 |
+
- Exadata Statistics (IO, Flash Cache, Smart Scan)
|
129 |
+
- Recommended Next Steps to Bridge Gaps
|
130 |
+
"""
|
131 |
+
if performance_test_mode and exadata_model and rack_size:
|
132 |
+
specs = exadata_specs.get(exadata_model, {}).get(rack_size, {})
|
133 |
+
if specs:
|
134 |
+
prompt += f"""
|
135 |
+
|
136 |
+
This was a PERFORMANCE TEST on Oracle Exadata {exadata_model} {rack_size}.
|
137 |
+
Theoretical Max:
|
138 |
+
- IOPS: {specs['max_iops']}
|
139 |
+
- Throughput: {specs['max_throughput']} GB/s
|
140 |
+
|
141 |
+
Compare observed vs theoretical. Recommend actions to close the performance gap.
|
142 |
+
"""
|
143 |
+
response = client.chat.completions.create(
|
144 |
+
model=llm_model,
|
145 |
+
messages=[
|
146 |
+
{"role": "system", "content": "You are an expert Oracle DBA."},
|
147 |
+
{"role": "user", "content": prompt}
|
148 |
+
]
|
149 |
+
)
|
150 |
+
return response.choices[0].message.content.strip()
|
151 |
+
|
152 |
+
class HealthAgent:
|
153 |
+
def check_health(self, content, llm_model):
|
154 |
+
cleaned_content = clean_awr_content(content)
|
155 |
+
if len(cleaned_content) > 128000:
|
156 |
+
cleaned_content = cleaned_content[:128000] + "\n\n[TRUNCATED]..."
|
157 |
+
|
158 |
+
prompt = f"""You are the Oracle AWR Health Analysis Agent.
|
159 |
+
|
160 |
+
Your primary responsibility is to detect and report ANY and ALL database health risks, alerts, warnings, or failures in the AWR report.
|
161 |
+
|
162 |
+
You MUST:
|
163 |
+
- Identify all issues marked as CRITICAL, WARNING, ALERT, FAILED, OFFLINE, CONFINED, DROPPED, or ERROR.
|
164 |
+
- Never omit or generalize. If something appears important, call it out.
|
165 |
+
- Classify each issue into: π¨ CRITICAL / β οΈ WARNING / β
INFO
|
166 |
+
- For CRITICAL and WARNING, provide suggested actions or considerations.
|
167 |
+
- Always confirm at the end if no CRITICAL or WARNING issues were found.
|
168 |
+
|
169 |
+
Special Attention Areas:
|
170 |
+
- Flash Cache or Flash Disk Failures
|
171 |
+
- I/O Subsystem stalls or errors
|
172 |
+
- ASM/Grid Disk issues
|
173 |
+
- Smart Scan failures
|
174 |
+
- Redo Log issues
|
175 |
+
- RAC Interconnect issues
|
176 |
+
|
177 |
+
AWR CONTENT:
|
178 |
+
{cleaned_content}
|
179 |
+
"""
|
180 |
+
response = client.chat.completions.create(
|
181 |
+
model=llm_model,
|
182 |
+
messages=[
|
183 |
+
{"role": "system", "content": "You are the strict Oracle AWR Health Analysis Agent."},
|
184 |
+
{"role": "user", "content": prompt}
|
185 |
+
]
|
186 |
+
)
|
187 |
+
return response.choices[0].message.content.strip()
|
188 |
+
|
189 |
+
class RaterAgent:
|
190 |
+
def rate(self, content):
|
191 |
+
prompt = f"Rate the following analysis from 1-5 stars and explain:\n\n{content}"
|
192 |
+
response = openrouter.chat.completions.create(
|
193 |
+
model="mistralai/Mixtral-8x7B-Instruct",
|
194 |
+
messages=[{"role": "user", "content": prompt}]
|
195 |
+
)
|
196 |
+
return response.choices[0].message.content.strip()
|
197 |
+
|
198 |
+
# === MAIN AWR PROCESS ===
|
199 |
+
def process_awr(awr_text, threshold, performance_test_mode, exadata_model, rack_size, llm_model):
|
200 |
+
analyzer = CriticalAnalyzerAgent()
|
201 |
+
health = HealthAgent()
|
202 |
+
rater = RaterAgent()
|
203 |
+
|
204 |
+
if not awr_text.strip():
|
205 |
+
return "No AWR text provided", "", "", ""
|
206 |
+
|
207 |
+
analysis = analyzer.analyze(awr_text, performance_test_mode, exadata_model, rack_size, llm_model)
|
208 |
+
health_status = health.check_health(awr_text, llm_model)
|
209 |
+
rating_text = rater.rate(analysis)
|
210 |
+
|
211 |
+
stars = 0
|
212 |
+
match = re.search(r"(\d+)", rating_text)
|
213 |
+
if match:
|
214 |
+
stars = int(match.group(1))
|
215 |
+
|
216 |
+
retry_status = "β
Accepted"
|
217 |
+
if stars < threshold:
|
218 |
+
analysis = analyzer.analyze(awr_text, performance_test_mode, exadata_model, rack_size, llm_model)
|
219 |
+
rating_text = rater.rate(analysis)
|
220 |
+
retry_status = "β
Retry Occurred"
|
221 |
+
|
222 |
+
return analysis, health_status, rating_text, retry_status
|
223 |
+
|
224 |
# === Gradio UI ===
|
225 |
with gr.Blocks() as demo:
|
226 |
with gr.Tab("Manual AWR Analysis"):
|
227 |
+
gr.Markdown("# π§ Multi-Agent Oracle AWR Analyzer (Production Edition)")
|
228 |
awr_text = gr.Textbox(label="Paste AWR Report", lines=30)
|
229 |
threshold = gr.Slider(0, 5, value=3, step=1, label="Correctness Threshold (Stars)")
|
230 |
performance_test_mode = gr.Checkbox(label="Performance Test Mode")
|
|
|
242 |
rating = gr.Textbox(label="Rater", lines=3)
|
243 |
retry_status = gr.Textbox(label="Retry Status")
|
244 |
|
|
|
245 |
analyze_btn.click(process_awr,
|
246 |
inputs=[awr_text, threshold, performance_test_mode, exadata_model, rack_size, llm_selector],
|
247 |
outputs=[output, health, rating, retry_status])
|
|
|
251 |
upload_status = gr.Textbox(label="Upload Status")
|
252 |
upload_file.upload(fn=upload_awr_file, inputs=upload_file, outputs=upload_status)
|
253 |
|
254 |
+
refresh_button = gr.Button("π Refresh File List")
|
255 |
file_multiselect = gr.Dropdown(choices=[], label="Select AWR Files", multiselect=True)
|
256 |
refresh_button.click(fn=lambda: gr.update(choices=list_awr_files()), outputs=file_multiselect)
|
257 |
|
|
|
261 |
|
262 |
if __name__ == "__main__":
|
263 |
demo.launch(debug=True)
|
264 |
+
|
265 |
+
|