Skip to content

Commit 8ae23ba

Browse files
committed
fix: update Gemini model configurations for better stability
- Changed model configurations for primary_agent, assistant, refiner, adviser, and pentester to use `gemini-2.5-pro`, enhancing performance with adjusted temperature and top_p values. - Updated pricing parameters for various models to reflect new pricing structures. - Revised the test report to include updated success rates and average latencies, ensuring accurate performance metrics for all agents. - Improved overall test results with enhanced latency measurements across multiple test cases.
1 parent cda16e2 commit 8ae23ba

File tree

2 files changed

+345
-335
lines changed

2 files changed

+345
-335
lines changed

backend/pkg/providers/gemini/config.yml

Lines changed: 29 additions & 19 deletions
Original file line numberDiff line numberDiff line change
@@ -21,6 +21,8 @@ simple_json:
2121

2222
primary_agent:
2323
model: gemini-2.5-flash
24+
temperature: 0.8
25+
top_p: 0.95
2426
n: 1
2527
max_tokens: 6000
2628
reasoning:
@@ -30,14 +32,16 @@ primary_agent:
3032
output: 2.5
3133

3234
assistant:
33-
model: gemini-2.5-flash
35+
model: gemini-2.5-pro
36+
temperature: 0.8
37+
top_p: 0.95
3438
n: 1
3539
max_tokens: 6000
3640
reasoning:
3741
effort: medium
3842
price:
39-
input: 0.3
40-
output: 2.5
43+
input: 1.25
44+
output: 10.0
4145

4246
generator:
4347
model: gemini-2.5-pro
@@ -52,28 +56,32 @@ generator:
5256
output: 10.0
5357

5458
refiner:
55-
model: gemini-2.5-flash
59+
model: gemini-2.5-pro
60+
temperature: 0.7
61+
top_p: 0.95
5662
n: 1
5763
max_tokens: 8000
5864
reasoning:
5965
effort: medium
6066
price:
61-
input: 0.3
62-
output: 2.5
67+
input: 1.25
68+
output: 10.0
6369

6470
adviser:
65-
model: gemini-2.5-flash
71+
model: gemini-2.5-pro
72+
temperature: 0.8
73+
top_p: 0.95
6674
n: 1
6775
max_tokens: 6000
6876
reasoning:
6977
effort: medium
7078
price:
71-
input: 0.3
72-
output: 2.5
79+
input: 1.25
80+
output: 10.0
7381

7482
reflector:
7583
model: gemini-2.0-flash
76-
temperature: 0.6
84+
temperature: 0.7
7785
top_p: 0.95
7886
n: 1
7987
max_tokens: 4000
@@ -83,7 +91,7 @@ reflector:
8391

8492
searcher:
8593
model: gemini-2.0-flash
86-
temperature: 0.6
94+
temperature: 0.7
8795
top_p: 0.95
8896
n: 1
8997
max_tokens: 5000
@@ -93,7 +101,7 @@ searcher:
93101

94102
enricher:
95103
model: gemini-2.0-flash
96-
temperature: 0.6
104+
temperature: 0.7
97105
top_p: 0.95
98106
n: 1
99107
max_tokens: 5000
@@ -114,23 +122,25 @@ coder:
114122
output: 10.0
115123

116124
installer:
117-
model: gemini-2.5-flash-lite
125+
model: gemini-2.5-flash
126+
temperature: 0.8
127+
top_p: 0.95
118128
n: 1
119129
max_tokens: 6000
120130
reasoning:
121131
effort: low
122132
price:
123-
input: 0.1
124-
output: 0.4
133+
input: 0.3
134+
output: 2.5
125135

126136
pentester:
127-
model: gemini-2.5-flash
137+
model: gemini-2.5-pro
128138
temperature: 0.8
129139
top_p: 0.95
130140
n: 1
131141
max_tokens: 6000
132142
reasoning:
133-
effort: medium
143+
effort: low
134144
price:
135-
input: 0.3
136-
output: 2.5
145+
input: 1.25
146+
output: 10.0

0 commit comments

Comments
 (0)