zidane 1.5.0 → 1.6.0

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/README.md CHANGED
@@ -35,11 +35,12 @@ All options on `createAgent`:
35
35
  createAgent({
36
36
  provider, // required: LLM provider
37
37
  harness: basic, // tool set (default: noTools)
38
- enableTools: true, // false for pure chat mode
39
- toolExecution: 'sequential', // or 'parallel'
40
- maxTurns: 50, // max loop iterations
41
- maxTokens: 16384, // max tokens per LLM response
42
- thinkingBudget: 10240, // exact thinking token budget
38
+ behavior: { // agent-level defaults
39
+ toolExecution: 'sequential', // or 'parallel'
40
+ maxTurns: 50, // max loop iterations
41
+ maxTokens: 16384, // max tokens per LLM response
42
+ thinkingBudget: 10240, // exact thinking token budget
43
+ },
43
44
  execution: createProcessContext(), // where tools run
44
45
  mcpServers: [], // MCP tool servers
45
46
  session, // session for persistence
@@ -51,19 +52,22 @@ All options on `agent.run()`:
51
52
 
52
53
  ```ts
53
54
  await agent.run({
54
- prompt: 'your task', // required
55
+ prompt: 'your task', // required
55
56
  model: 'claude-opus-4-6',
56
57
  system: 'be concise',
57
- thinking: 'medium', // off | minimal | low | medium | high
58
- thinkingBudget: 8192, // overrides level-based default
59
- maxTurns: 10, // overrides agent-level default
60
- maxTokens: 4096, // overrides agent-level default
61
- images: [], // base64 images
58
+ thinking: 'medium', // off | minimal | low | medium | high
59
+ behavior: { // per-run overrides
60
+ maxTurns: 10,
61
+ maxTokens: 4096,
62
+ thinkingBudget: 8192,
63
+ },
64
+ tools: {}, // override tools for this run ({} = no tools)
65
+ images: [], // base64 images
62
66
  signal: abortController.signal,
63
67
  })
64
68
  ```
65
69
 
66
- Per-run options override agent-level defaults. Agent-level defaults override hardcoded defaults.
70
+ Precedence: `run.behavior` > `agent.behavior` > `harness.behavior` > hardcoded defaults.
67
71
 
68
72
  ## CLI
69
73
 
@@ -139,10 +143,10 @@ const harness = defineHarness({
139
143
  })
140
144
  ```
141
145
 
142
- For pure chat with no tools:
146
+ For pure chat with no tools, pass `tools: {}` on a specific run or use the `noTools` harness:
143
147
 
144
148
  ```ts
145
- const agent = createAgent({ provider, enableTools: false })
149
+ await agent.run({ prompt: 'just chat', tools: {} })
146
150
  ```
147
151
 
148
152
  ## Thinking