loki-mode 5.2.3 → 5.3.0

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
@@ -25,6 +25,16 @@
25
25
 
26
26
  ---
27
27
 
28
+ ## Presentation
29
+
30
+ ![Loki Mode Presentation](docs/loki-mode-presentation.gif)
31
+
32
+ *9 slides: Problem, Solution, 37 Agents, RARV Cycle, Benchmarks, Multi-Provider, Full Lifecycle*
33
+
34
+ **[Download PPTX](docs/loki-mode-presentation.pptx)** for offline viewing
35
+
36
+ ---
37
+
28
38
  ## Usage
29
39
 
30
40
  ### Option 1: Claude Code Skill (Recommended)
package/SKILL.md CHANGED
@@ -3,7 +3,7 @@ name: loki-mode
3
3
  description: Multi-agent autonomous startup system. Triggers on "Loki Mode". Takes PRD to deployed product with zero human intervention. Requires --dangerously-skip-permissions flag.
4
4
  ---
5
5
 
6
- # Loki Mode v5.2.3
6
+ # Loki Mode v5.3.0
7
7
 
8
8
  **You are an autonomous agent. You make decisions. You do not ask questions. You do not stop.**
9
9
 
@@ -88,15 +88,17 @@ These rules are ABSOLUTE. Violating them is a critical failure.
88
88
 
89
89
  ## Model Selection
90
90
 
91
- | Task Type | Tier | Claude | Codex | Gemini |
92
- |-----------|------|--------|-------|--------|
93
- | PRD analysis, architecture, system design | **planning** | opus | effort=xhigh | thinking=high |
94
- | Feature implementation, complex bugs | **development** | sonnet | effort=high | thinking=medium |
95
- | Code review (always 3 parallel reviewers) | **development** | sonnet | effort=high | thinking=medium |
96
- | Integration tests, E2E, deployment | **development** | sonnet | effort=high | thinking=medium |
97
- | Unit tests, linting, docs, simple fixes | **fast** | haiku | effort=low | thinking=low |
91
+ **Default (v5.3.0):** Haiku disabled for quality. Use `--allow-haiku` or `LOKI_ALLOW_HAIKU=true` to enable.
98
92
 
99
- **Parallelization rule (Claude only):** Launch up to 10 haiku agents simultaneously for independent tasks.
93
+ | Task Type | Tier | Claude (default) | Claude (--allow-haiku) | Codex | Gemini |
94
+ |-----------|------|------------------|------------------------|-------|--------|
95
+ | PRD analysis, architecture, system design | **planning** | opus | opus | effort=xhigh | thinking=high |
96
+ | Feature implementation, complex bugs | **development** | opus | sonnet | effort=high | thinking=medium |
97
+ | Code review (always 3 parallel reviewers) | **development** | opus | sonnet | effort=high | thinking=medium |
98
+ | Integration tests, E2E, deployment | **development** | opus | sonnet | effort=high | thinking=medium |
99
+ | Unit tests, linting, docs, simple fixes | **fast** | sonnet | haiku | effort=low | thinking=low |
100
+
101
+ **Parallelization rule (Claude only):** Launch up to 10 agents simultaneously for independent tasks.
100
102
 
101
103
  **Degraded mode (Codex/Gemini):** No parallel agents or Task tool. Runs RARV cycle sequentially. See `skills/model-selection.md`.
102
104
 
@@ -227,4 +229,4 @@ Auto-detected or force with `LOKI_COMPLEXITY`:
227
229
 
228
230
  ---
229
231
 
230
- **v5.2.3 | CoVe + MemEvolve + Quality Gates | ~230 lines core**
232
+ **v5.3.0 | Haiku Control Flag + Quality Model Defaults | ~240 lines core**
package/VERSION CHANGED
@@ -1 +1 @@
1
- 5.2.3
1
+ 5.3.0
package/autonomy/run.sh CHANGED
@@ -53,6 +53,12 @@
53
53
  # LOKI_PERPETUAL_MODE - Ignore ALL completion signals (default: false)
54
54
  # Set to 'true' for truly infinite operation
55
55
  #
56
+ # Model Selection:
57
+ # LOKI_ALLOW_HAIKU - Enable Haiku model for fast tier (default: false)
58
+ # When false: Opus for dev/bugfix, Sonnet for tests/docs
59
+ # When true: Sonnet for dev, Haiku for tests/docs (original)
60
+ # Use --allow-haiku flag or set to 'true'
61
+ #
56
62
  # 2026 Research Enhancements:
57
63
  # LOKI_PROMPT_REPETITION - Enable prompt repetition for Haiku agents (default: true)
58
64
  # arXiv 2512.14982v1: Improves accuracy 4-5x on structured tasks
@@ -3867,6 +3873,11 @@ main() {
3867
3873
  PARALLEL_MODE=true
3868
3874
  shift
3869
3875
  ;;
3876
+ --allow-haiku)
3877
+ export LOKI_ALLOW_HAIKU=true
3878
+ log_info "Haiku model enabled for fast tier"
3879
+ shift
3880
+ ;;
3870
3881
  --provider)
3871
3882
  if [[ -n "${2:-}" ]]; then
3872
3883
  LOKI_PROVIDER="$2"
@@ -3909,6 +3920,7 @@ main() {
3909
3920
  echo ""
3910
3921
  echo "Options:"
3911
3922
  echo " --parallel Enable git worktree-based parallel workflows"
3923
+ echo " --allow-haiku Enable Haiku model for fast tier (default: disabled)"
3912
3924
  echo " --provider <name> Provider: claude (default), codex, gemini"
3913
3925
  echo " --help, -h Show this help message"
3914
3926
  echo ""
@@ -13,7 +13,7 @@ const skillDir = path.join(homeDir, '.claude', 'skills', 'loki-mode');
13
13
  const packageDir = path.join(__dirname, '..');
14
14
 
15
15
  console.log('');
16
- console.log('Loki Mode v5.2.3 installed!');
16
+ console.log('Loki Mode v5.2.4 installed!');
17
17
  console.log('');
18
18
 
19
19
  // Try to create skill symlink
Binary file
Binary file
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "loki-mode",
3
- "version": "5.2.3",
3
+ "version": "5.3.0",
4
4
  "description": "Multi-agent autonomous startup system for Claude Code, Codex CLI, and Gemini CLI",
5
5
  "keywords": [
6
6
  "claude",
@@ -44,13 +44,23 @@ PROVIDER_HAS_MCP=true
44
44
  PROVIDER_MAX_PARALLEL=10
45
45
 
46
46
  # Model Configuration (Abstract Tiers)
47
+ # Default: Haiku disabled for quality. Use --allow-haiku or LOKI_ALLOW_HAIKU=true to enable.
47
48
  PROVIDER_MODEL_PLANNING="claude-opus-4-5-20251101"
48
- PROVIDER_MODEL_DEVELOPMENT="claude-sonnet-4-5-20251101"
49
- PROVIDER_MODEL_FAST="claude-haiku-4-5-20251101"
49
+ PROVIDER_MODEL_DEVELOPMENT="claude-opus-4-5-20251101" # Opus for dev (was sonnet)
50
+ if [ "${LOKI_ALLOW_HAIKU:-false}" = "true" ]; then
51
+ PROVIDER_MODEL_FAST="claude-haiku-4-5-20251101"
52
+ PROVIDER_MODEL_DEVELOPMENT="claude-sonnet-4-5-20251101" # Sonnet for dev when haiku enabled
53
+ else
54
+ PROVIDER_MODEL_FAST="claude-sonnet-4-5-20251101" # Sonnet for fast (no haiku)
55
+ fi
50
56
 
51
57
  # Model Selection (for Task tool)
52
58
  PROVIDER_TASK_MODEL_PARAM="model"
53
- PROVIDER_TASK_MODEL_VALUES=("opus" "sonnet" "haiku")
59
+ if [ "${LOKI_ALLOW_HAIKU:-false}" = "true" ]; then
60
+ PROVIDER_TASK_MODEL_VALUES=("opus" "sonnet" "haiku")
61
+ else
62
+ PROVIDER_TASK_MODEL_VALUES=("opus" "sonnet") # No haiku option
63
+ fi
54
64
 
55
65
  # Context and Limits
56
66
  PROVIDER_CONTEXT_WINDOW=200000
@@ -87,14 +97,28 @@ provider_invoke() {
87
97
  }
88
98
 
89
99
  # Model tier to Task tool model parameter value
100
+ # Respects LOKI_ALLOW_HAIKU flag for tier mapping
90
101
  provider_get_tier_param() {
91
102
  local tier="$1"
92
- case "$tier" in
93
- planning) echo "opus" ;;
94
- development) echo "sonnet" ;;
95
- fast) echo "haiku" ;;
96
- *) echo "sonnet" ;; # default to development tier
97
- esac
103
+ if [ "${LOKI_ALLOW_HAIKU:-false}" = "true" ]; then
104
+ # With haiku: original tier mapping
105
+ case "$tier" in
106
+ planning) echo "opus" ;;
107
+ development) echo "sonnet" ;;
108
+ fast) echo "haiku" ;;
109
+ *) echo "sonnet" ;;
110
+ esac
111
+ else
112
+ # Without haiku (default): upgrade all tiers
113
+ # - Development + bug fixes -> opus
114
+ # - Testing + documentation -> sonnet
115
+ case "$tier" in
116
+ planning) echo "opus" ;;
117
+ development) echo "opus" ;; # Upgraded from sonnet
118
+ fast) echo "sonnet" ;; # Upgraded from haiku
119
+ *) echo "opus" ;; # Default to opus
120
+ esac
121
+ fi
98
122
  }
99
123
 
100
124
  # Tier-aware invocation (Claude supports model selection via --model flag)
@@ -20,11 +20,33 @@ Loki Mode supports three AI providers. Claude has full features; Codex and Gemin
20
20
 
21
21
  ## Abstract Model Tiers
22
22
 
23
- | Tier | Purpose | Claude | Codex | Gemini |
24
- |------|---------|--------|-------|--------|
25
- | **planning** | PRD analysis, architecture, system design | opus | effort=xhigh | thinking=high |
26
- | **development** | Feature implementation, complex bugs, tests | sonnet | effort=high | thinking=medium |
27
- | **fast** | Unit tests, docs, linting, simple tasks | haiku | effort=low | thinking=low |
23
+ **Default (v5.3.0):** Haiku disabled for quality. All tasks use Opus or Sonnet.
24
+
25
+ | Tier | Purpose | Claude (default) | Claude (--allow-haiku) | Codex | Gemini |
26
+ |------|---------|------------------|------------------------|-------|--------|
27
+ | **planning** | PRD analysis, architecture, system design | opus | opus | effort=xhigh | thinking=high |
28
+ | **development** | Feature implementation, complex bugs, tests | opus | sonnet | effort=high | thinking=medium |
29
+ | **fast** | Unit tests, docs, linting, simple tasks | sonnet | haiku | effort=low | thinking=low |
30
+
31
+ ### Enabling Haiku
32
+
33
+ To enable Haiku for fast-tier tasks (cost optimization at potential quality trade-off):
34
+
35
+ ```bash
36
+ # Via environment variable
37
+ LOKI_ALLOW_HAIKU=true ./autonomy/run.sh ./prd.md
38
+
39
+ # Via CLI flag
40
+ ./autonomy/run.sh --allow-haiku ./prd.md
41
+
42
+ # Via loki CLI wrapper
43
+ loki start --allow-haiku ./prd.md
44
+ ```
45
+
46
+ When Haiku is enabled:
47
+ - Development tier uses Sonnet (instead of Opus)
48
+ - Fast tier uses Haiku (instead of Sonnet)
49
+ - Planning tier always uses Opus (unchanged)
28
50
 
29
51
  ---
30
52
 
@@ -36,7 +58,7 @@ Loki Mode supports three AI providers. Claude has full features; Codex and Gemin
36
58
  | **development** | Development, QA, Deployment | Feature implementation, complex bugs, integration/E2E tests, code review, deployment |
37
59
  | **fast** | All other operations (parallel for Claude) | Unit tests, docs, bash commands, linting, monitoring |
38
60
 
39
- **Claude-specific model names:** opus, sonnet, haiku
61
+ **Claude-specific model names:** opus, sonnet, haiku (haiku requires --allow-haiku flag)
40
62
  **Codex effort levels:** xhigh, high, medium, low
41
63
  **Gemini thinking levels:** high, medium, low
42
64