phi-core 0.7.0

Simple, effective agent loop with tool execution and event streaming
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92

# Full coding agent configuration
# Demonstrates agent profile, provider, tools, compaction, and execution limits.

# ── Agent profile ───────────────────────────────────────────────────────────

[agent.profile]
name = "coding-agent"
description = "A coding assistant with file and shell access"
system_prompt = """You are an expert software engineer. You read code carefully,
make minimal targeted changes, and always verify your work by running tests."""
thinking_level = "high"       # "off", "minimal", "low", "medium", "high"
temperature = 0.2             # Low temperature for precise code generation
max_tokens = 16384            # Max output tokens per response

# ── Profile instances (named variations of the profile blueprint) ──────────

[[agent.profile.instances]]
id = "{{%coder%}}"
description = "A code generation specialist"
thinking_level = "high"
temperature = 0.2
max_tokens = 16384
config_id = "coder"

[[agent.profile.instances]]
id = "{{%reviewer%}}"
description = "A code review specialist"
thinking_level = "high"
temperature = 0.1
max_tokens = 8192
config_id = "reviewer"

# ── Agent instances (reference profile instances) ──────────────────────────

[[agent.instances]]
name = "code-writer"
agent_profile = "{{agent_profile.coder}}"
system_prompt = "You write clean, well-tested code. Follow existing patterns."

[[agent.instances]]
name = "code-reviewer"
agent_profile = "{{agent_profile.reviewer}}"
system_prompt = "You review code for bugs, security issues, and style violations."

# ── Provider ────────────────────────────────────────────────────────────────

[provider]
model = "claude-sonnet-4-20250514"
api_key = "${ANTHROPIC_API_KEY}"
api = "anthropic_messages"    # Explicit protocol (auto-detected from model if omitted)

[provider.cost]
input_per_million = 3.0       # Cost rates for budget tracking
output_per_million = 15.0
cache_read_per_million = 0.3
cache_write_per_million = 3.75

# ── Tools ───────────────────────────────────────────────────────────────────

[tools]
# Tool names resolved by the caller's tool registry at runtime
enabled = ["bash", "file_read", "file_write", "search"]
tool_strategy = "parallel"    # "sequential", "parallel", or "batched"

# ── Compaction ──────────────────────────────────────────────────────────────

[compaction]
max_context_tokens = 200000   # Model context window
system_prompt_tokens = 4000   # Tokens reserved for system prompt
compact_at_pct = 0.85         # Start compaction at 85% capacity
keep_first_turns = 2          # Always keep first 2 turns verbatim
keep_recent_turns = 4         # Keep last 4 turns verbatim
max_summary_tokens = 8000     # Budget for summarised middle section
tool_output_max_lines = 100   # Truncate tool output in kept turns

# ── Execution limits ────────────────────────────────────────────────────────

[execution]
max_turns = 50                # Maximum LLM turns per loop
max_total_tokens = 1000000    # Total token budget across all turns
max_duration_secs = 600       # 10-minute wall-clock limit
max_cost = 5.0                # $5 cost cap

[execution.retry]
max_retries = 3               # Retry transient provider errors
initial_delay_ms = 1000       # 1 second initial delay
backoff_multiplier = 2.0      # Exponential backoff
max_delay_ms = 30000          # 30 second max delay

[execution.cache]
enabled = true
strategy = "auto"             # "auto", "disabled"