gyoshu 0.2.0 → 0.2.2

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/AGENTS.md CHANGED
@@ -18,6 +18,46 @@ Gyoshu is a scientific research agent extension for OpenCode. It provides:
18
18
  | **Baksa** | PhD Reviewer | 박사 | Adversarial verifier - challenges claims, calculates trust scores |
19
19
  | **Jogyo Paper Writer** | Grad Student | 조교 | Transforms raw findings into narrative research reports |
20
20
 
21
+ ## Model Configuration
22
+
23
+ Gyoshu uses free OpenCode models by default for zero-configuration setup:
24
+
25
+ | Agent | Default Model | Role |
26
+ |-------|---------------|------|
27
+ | **Gyoshu** | `opencode/glm-4.7-free` | Research planner |
28
+ | **Baksa** | `opencode/alpha-minimax-m2` | Adversarial verifier |
29
+ | **Jogyo** | `opencode/grok-code` | Research executor |
30
+ | **Jogyo Paper Writer** | `opencode/grok-code` | Report writer |
31
+ | **Jogyo Feedback** | `opencode/grok-code` | Feedback explorer |
32
+ | **Jogyo Insight** | `opencode/grok-code` | Evidence gatherer |
33
+
34
+ ### Recommended Models for Maximum Performance
35
+
36
+ For best research quality, we recommend using Anthropic's Claude models:
37
+
38
+ | Agent | Recommended Model | Why |
39
+ |-------|-------------------|-----|
40
+ | **Gyoshu** | `anthropic/claude-opus-4-5-high` | Complex research planning requires top-tier reasoning |
41
+ | **Baksa** | `anthropic/claude-opus-4-5-high` | Adversarial verification needs strong critical thinking |
42
+ | **Jogyo** | `anthropic/claude-sonnet-4-5-high` | Balanced capability for code execution |
43
+ | **Jogyo subagents** | `anthropic/claude-sonnet-4-5-high` | Consistent quality across tasks |
44
+
45
+ ### Changing Models
46
+
47
+ To change an agent's model, edit the `model:` field in the YAML frontmatter of `src/agent/{agent-name}.md`:
48
+
49
+ ```yaml
50
+ ---
51
+ mode: subagent
52
+ description: ...
53
+ model: anthropic/claude-sonnet-4-5-high # Change this line
54
+ temperature: 0.3
55
+ ...
56
+ ---
57
+ ```
58
+
59
+ > **Note**: The default free models work out-of-the-box. Premium models (Anthropic, OpenAI) require API keys configured in OpenCode.
60
+
21
61
  ## Build & Test Commands
22
62
 
23
63
  ### Python Tests (pytest)
package/package.json CHANGED
@@ -1,10 +1,10 @@
1
1
  {
2
2
  "name": "gyoshu",
3
- "version": "0.2.0",
3
+ "version": "0.2.2",
4
4
  "description": "Scientific research agent extension for OpenCode - turns research goals into reproducible Jupyter notebooks",
5
5
  "type": "module",
6
6
  "bin": {
7
- "gyoshu": "./bin/gyoshu.js"
7
+ "gyoshu": "bin/gyoshu.js"
8
8
  },
9
9
  "files": [
10
10
  "bin/",
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  mode: subagent
3
3
  description: Adversarial PhD reviewer that challenges Jogyo's research claims and verifies evidence
4
- model: openai/gpt-5.2-xhigh
4
+ model: opencode/alpha-minimax-m2
5
5
  temperature: 0.3
6
6
  maxSteps: 15
7
7
  tools:
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  mode: primary
3
3
  description: Scientific research planner - orchestrates research workflows and manages REPL lifecycle
4
- model: anthropic/claude-opus-4-5-high
4
+ model: opencode/glm-4.7-free
5
5
  temperature: 0.3
6
6
  maxSteps: 50
7
7
  tools:
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  mode: subagent
3
3
  description: Explores retrospective feedback to extract lessons and patterns for research improvement
4
- model: anthropic/claude-sonnet-4-5-high
4
+ model: opencode/grok-code
5
5
  temperature: 0.3
6
6
  maxSteps: 10
7
7
  tools:
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  mode: subagent
3
3
  description: Gathers evidence from previous notebooks, URLs, and documentation for research support
4
- model: anthropic/claude-sonnet-4-5-high
4
+ model: opencode/grok-code
5
5
  temperature: 0.3
6
6
  maxSteps: 15
7
7
  tools:
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  mode: subagent
3
3
  description: Generates human-readable, narrative research reports from structured context
4
- model: anthropic/claude-sonnet-4-5-high
4
+ model: opencode/grok-code
5
5
  temperature: 0.4
6
6
  maxSteps: 5
7
7
  tools:
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  mode: subagent
3
3
  description: Scientific research agent with Python REPL and structured output markers
4
- model: anthropic/claude-sonnet-4-5-high
4
+ model: opencode/grok-code
5
5
  temperature: 0.2
6
6
  maxSteps: 50
7
7
  tools: