power / gui /config.py
Deepakresearcher's picture
New build
b555842
Raw History Blame Contribute Delete
19.5 kB
"""
gui/config.py
─────────────────────────────────────────────────────────────────────────────
Central configuration for A-LEMS.
3-layer navigation:
Layer 1 — Sidebar: 11 section buttons (SECTIONS)
Layer 2 — Section landing: card grid (SECTION_PAGES)
Layer 3 — Page: actual page module
Session state keys (set by sidebar.py, read by streamlit_app.py):
nav_section — active section name | None → None means Overview
nav_page — active page_id | None → None means section landing
nav_last — dict[section → last page_id visited] (resume chip)
─────────────────────────────────────────────────────────────────────────────
"""
from pathlib import Path
import yaml
# ── Paths ─────────────────────────────────────────────────────────────────────
PROJECT_ROOT = Path(__file__).parent.parent
DB_PATH = PROJECT_ROOT / "data" / "experiments.db"
LIVE_API = "http://localhost:8765"
CONFIG_DIR = PROJECT_ROOT / "config"
# ── Human-readable energy comparisons ────────────────────────────────────────
_PHONE_CHARGE_J = 20000
_WHATSAPP_MSG_J = 0.014
_GOOGLE_SEARCH_J = 0.3
_BABY_FEED_ML = 200
_CO2_TREE_SEQ_KG_PER_YEAR = 22
_WATER_BOTTLE_ML = 500
# ── YAML config loader ────────────────────────────────────────────────────────
def _load_yaml(filename: str) -> dict:
path = CONFIG_DIR / filename
try:
with open(path) as f:
return yaml.safe_load(f) or {}
except FileNotFoundError:
return {}
except Exception as e:
print(f"[config] Warning: could not load {filename}: {e}")
return {}
INSIGHTS_RULES = _load_yaml("insights_rules.yaml")
DASHBOARD_CFG = _load_yaml("dashboard.yaml")
DESIGNER_CFG = _load_yaml("experiment_designer.yaml")
TEMPLATES_CFG = _load_yaml("experiment_templates.yaml")
GAP_RULES = _load_yaml("gap_detection.yaml")
# ── Plotly dark theme ─────────────────────────────────────────────────────────
PL = dict(
paper_bgcolor="#0f1520", plot_bgcolor="#090d13",
font=dict(family="IBM Plex Mono, monospace", size=10, color="#7090b0"),
margin=dict(l=40, r=20, t=30, b=30),
legend=dict(bgcolor="rgba(0,0,0,0)", font=dict(size=9)),
colorway=["#22c55e", "#ef4444", "#3b82f6", "#f59e0b", "#38bdf8", "#a78bfa"],
xaxis=dict(gridcolor="#1e2d45", linecolor="#1e2d45", tickfont=dict(size=9)),
yaxis=dict(gridcolor="#1e2d45", linecolor="#1e2d45", tickfont=dict(size=9)),
)
# ── Colour maps ───────────────────────────────────────────────────────────────
WF_COLORS = {"linear": "#22c55e", "agentic": "#ef4444"}
STATUS_COLORS = {
"completed": "#3b82f6",
"running": "#22c55e",
"pending": "#f59e0b",
"not_started": "#4b5563",
"failed": "#ef4444",
"partial": "#f97316",
}
STATUS_ICONS = {
"completed": "●",
"running": "🟢",
"pending": "🟡",
"not_started": "○",
"failed": "🔴",
"partial": "◑",
}
# ── Section accent colours ────────────────────────────────────────────────────
SECTION_ACCENTS = {
"COMMAND CENTRE": "#22c55e",
"ENERGY & SILICON": "#f59e0b",
"AGENTIC INTELLIGENCE": "#ef4444",
"DATA MOVEMENT": "#a78bfa",
"SESSIONS & RUNS": "#3b82f6",
"RESEARCH & INSIGHTS": "#38bdf8",
"ENVIRONMENT": "#34d399",
"DATA QUALITY": "#f472b6",
"SILICON LAB": "#fb923c",
"DEVELOPER TOOLS": "#94a3b8",
"SETTINGS": "#475569",
}
# ── Page status constants ─────────────────────────────────────────────────────
EXISTS = "exists"
NEW = "new"
PLANNED = "planned"
BLOCKED = "blocked"
# ── SECTION_PAGES ─────────────────────────────────────────────────────────────
# Single source of truth for all navigation metadata.
# Used by: section_landing.py, streamlit_app.py, sidebar.py
#
# Section keys: accent, icon, who, description, pages[]
# Page keys: id, label, icon, desc, status, blocked_reason (optional)
#
SECTION_PAGES = {
"COMMAND CENTRE": {
"accent": "#22c55e", "icon": "◈",
"who": "All personas — start every session here",
"description": "Live session state, KPIs, experiment control, and multi-host dispatch.",
"pages": [
{"id": "overview", "label": "Overview", "icon": "◈", "desc": "Live session KPIs, energy summary, data sufficiency banner.", "status": EXISTS},
{"id": "silicon_journey", "label": "Ask Me", "icon": "◑", "desc": "Chat-style experiment — run on silicon, see energy, get insights.", "status": NEW},
{"id": "execute", "label": "Execute Run", "icon": "▶", "desc": "Launch experiments against local or remote lab targets.", "status": EXISTS},
{"id": "designer", "label": "Experiment Designer", "icon": "🧪", "desc": "Design conditional multi-container experiment plans.", "status": EXISTS},
{"id": "experiments", "label": "Experiments", "icon": "≡", "desc": "Browse all experiments with status, runs, and group metadata.", "status": EXISTS},
{"id": "multi_host_dispatch", "label": "Multi-Host Dispatch", "icon": "⊕", "desc": "Trigger parallel runs across N silicon targets simultaneously.", "status": PLANNED},
],
},
"ENERGY & SILICON": {
"accent": "#f59e0b", "icon": "⚡",
"who": "Silicon developers · chip engineers · CPU/GPU vendors",
"description": "RAPL domain breakdown, C-state residency, thermal behaviour, and idle baseline drift.",
"pages": [
{"id": "energy", "label": "Energy Lab", "icon": "⚡", "desc": "pkg / core / uncore / dram RAPL energy per run, power over time.", "status": EXISTS},
{"id": "cpu", "label": "CPU & C-States", "icon": "▣", "desc": "Core frequency, C-state residency, IPC, ring bus, context switches.", "status": EXISTS},
{"id": "thermal", "label": "Thermal Analysis", "icon": "🌡", "desc": "Throttle events, per-run temperature time series, thermal delta.", "status": NEW},
{"id": "baseline", "label": "Baseline & Idle", "icon": "⊟", "desc": "Idle baseline drift, governor state, turbo on/off energy impact.", "status": NEW},
{"id": "anomalies", "label": "Anomalies", "icon": "⚠", "desc": "Statistical outliers across all energy and performance metrics.", "status": EXISTS},
],
},
"AGENTIC INTELLIGENCE": {
"accent": "#ef4444", "icon": "⇌",
"who": "Orchestration engineers · MLOps · framework developers",
"description": "True cost of agentic overhead — planning, execution, synthesis phases, and orchestration tax.",
"pages": [
{"id": "agentic_linear", "label": "Agentic vs Linear", "icon": "⇌", "desc": "Side-by-side energy: agentic overhead vs linear baseline.", "status": EXISTS},
{"id": "tax", "label": "Tax Attribution", "icon": "▲", "desc": "Orchestration tax breakdown — what percentage is pure overhead.", "status": EXISTS},
{"id": "models", "label": "Apple-to-Apple", "icon": "🍎", "desc": "Controlled model comparison — same task, different models, same hardware.", "status": EXISTS},
{"id": "phase_drilldown", "label": "Phase Drilldown", "icon": "⬡", "desc": "Per-step planning/execution/synthesis energy from orchestration_events.", "status": NEW},
],
},
"DATA MOVEMENT": {
"accent": "#a78bfa", "icon": "◧",
"who": "Infrastructure engineers · cloud architects",
"description": "Energy cost per data layer — cache, memory, network, tokens, swap, interrupts.",
"pages": [
{"id": "data_cache", "label": "Cache & Memory", "icon": "◧", "desc": "Cache miss rate, rss/vms memory, IPC correlation with energy.", "status": NEW},
{"id": "data_tokens", "label": "Token Flow", "icon": "⟳", "desc": "LLM token movement, prompt/completion cost, latency per step.", "status": NEW},
{"id": "data_network", "label": "Network & API", "icon": "⌁", "desc": "DNS and API latency, idle energy cost of waiting for responses.", "status": NEW},
{"id": "data_swap", "label": "Swap & Paging", "icon": "⇅", "desc": "Swap usage, memory pressure, energy impact of paging activity.", "status": NEW},
{"id": "data_interrupts", "label": "Interrupts", "icon": "∿", "desc": "Interrupt rate time series, wakeup latency, per-run interrupt cost.", "status": NEW},
{"id": "data_page_faults", "label": "Page Faults", "icon": "⊘", "desc": "Major/minor page fault energy impact.", "status": BLOCKED,
"blocked_reason": "Fix PF-1 — page faults collected but not inserted into the runs table."},
{"id": "data_network_bytes", "label": "Network Bytes", "icon": "⊞", "desc": "bytes_sent / bytes_recv, TCP retransmits, effective throughput.", "status": BLOCKED,
"blocked_reason": "Collect bytes_sent / bytes_recv — not yet captured by the measurement harness."},
],
},
"SESSIONS & RUNS": {
"accent": "#3b82f6", "icon": "⬡",
"who": "All personas — browse, filter, compare, deep-dive",
"description": "Session management, run comparison, and single-run sensor stream drilldown.",
"pages": [
{"id": "sessions", "label": "Sessions", "icon": "⬡", "desc": "All sessions with metadata, status, experiment count, run totals.", "status": EXISTS},
{"id": "session_analysis", "label": "Session Analysis", "icon": "◑", "desc": "Deep statistical analysis of a selected session — trends, regression.", "status": EXISTS},
{"id": "explorer", "label": "Run Explorer", "icon": "⊞", "desc": "Filter and compare runs across experiments, models, and task types.", "status": EXISTS},
{"id": "run_drilldown", "label": "Run Drilldown", "icon": "🔍", "desc": "All sensor streams for one run — energy, CPU, thermal, interrupts.", "status": NEW},
],
},
"RESEARCH & INSIGHTS": {
"accent": "#38bdf8", "icon": "🔬",
"who": "Domain researchers · AI scientists · ML engineers",
"description": "Cross-run analysis, efficiency ratios, domain energy mapping, and ML feature workspace.",
"pages": [
{"id": "research_insights", "label": "Research Insights", "icon": "🔬", "desc": "Automated insight generation across all collected runs.", "status": EXISTS},
{"id": "query_analysis", "label": "Query Analysis", "icon": "◑", "desc": "Query class clustering, energy per query type, complexity scaling.", "status": EXISTS},
{"id": "domains", "label": "Domains", "icon": "◉", "desc": "Task domain energy profiles — coding vs writing vs reasoning vs math.", "status": EXISTS},
{"id": "scheduler", "label": "Scheduler", "icon": "〜", "desc": "Run scheduling analysis, queue length, kernel/user time split.", "status": EXISTS},
{"id": "efficiency", "label": "Efficiency Explorer","icon": "⚡", "desc": "energy_per_token, J/instruction, IPC correlation — find optimal configs.", "status": NEW},
{"id": "ml_features", "label": "ML Features", "icon": "⊟", "desc": "70-column ml_features view — correlation matrix, export for training.", "status": NEW},
{"id": "hypotheses", "label": "Hypothesis Tracker", "icon": "💡", "desc": "State research hypotheses, track supporting and contradicting evidence.", "status": NEW},
],
},
"ENVIRONMENT": {
"accent": "#34d399", "icon": "♻",
"who": "Sustainability teams · policy researchers · CTO offices",
"description": "Carbon, water, and methane cost of AI workloads by model, task, and geography.",
"pages": [
{"id": "sustainability", "label": "Sustainability", "icon": "♻", "desc": "Full environmental footprint per run — carbon, water, methane.", "status": EXISTS},
{"id": "carbon_country", "label": "Carbon by Country", "icon": "🌍", "desc": "Grid intensity × country_code — carbon cost mapped by geography.", "status": NEW},
{"id": "water_methane", "label": "Water & Methane", "icon": "💧", "desc": "Separate water_ml and methane_mg visualisation and trends.", "status": NEW},
],
},
"DATA QUALITY": {
"accent": "#f472b6", "icon": "✓",
"who": "All personas — trust your data before you analyse it",
"description": "Run validity, sensor completeness, statistical sufficiency, and hash integrity.",
"pages": [
{"id": "dq_validity", "label": "Run Validity", "icon": "✓", "desc": "Flags invalid runs — thermal throttle, bad baselines, experiment_valid.", "status": NEW},
{"id": "dq_coverage", "label": "Sensor Coverage", "icon": "◫", "desc": "NULL-count heatmap across 80+ columns — which sensors drop out when.", "status": NEW},
{"id": "dq_sufficiency", "label": "Sufficiency Advisor", "icon": "◈", "desc": "How many more experiments needed for statistical significance per cell.", "status": NEW},
{"id": "dq_integrity", "label": "Hash Integrity", "icon": "#", "desc": "run_state_hash verification — detect corrupted or tampered run records.", "status": NEW},
{"id": "dq_schema", "label": "Schema Log", "icon": "≡", "desc": "Schema version history — migration log and applied changes.", "status": NEW},
],
},
"SILICON LAB": {
"accent": "#fb923c", "icon": "▣",
"who": "Silicon developers · cloud engineers · HPC teams",
"description": "Hardware profiles, cross-silicon comparison, and multi-host experiment infrastructure.",
"pages": [
{"id": "hw_registry", "label": "Hardware Registry", "icon": "▣", "desc": "All hw_id records — full CPU spec, AVX/VMX flags, RAPL capabilities.", "status": NEW},
{"id": "silicon_compare", "label": "Cross-Silicon Compare","icon": "⇌", "desc": "Same experiment, different chips — controlled silicon comparison.", "status": NEW},
{"id": "multi_host_status", "label": "Multi-Host Status", "icon": "⊕", "desc": "Per-host connection health, tunnel state, last run timestamp.", "status": PLANNED},
{"id": "capability_matrix", "label": "Capability Matrix", "icon": "◫", "desc": "AVX2/AVX512/VMX/RAPL flags across all registered hardware.", "status": PLANNED},
],
},
"DEVELOPER TOOLS": {
"accent": "#94a3b8", "icon": "📋",
"who": "Engineers and developers building on A-LEMS",
"description": "Schema docs, SQL console, run replay, LLM call log, environment fingerprint, ML export.",
"pages": [
{"id": "live", "label": "Run Replay", "icon": "📼", "desc": "Post-run replay of energy and CPU samples with time scrubbing.", "status": EXISTS},
{"id": "schema_docs","label": "Schema & Docs", "icon": "📋", "desc": "Full database schema documentation with column descriptions.", "status": EXISTS},
{"id": "sql_query", "label": "SQL Query", "icon": "🔍", "desc": "Raw SQL console against the experiments database.", "status": EXISTS},
{"id": "env_config", "label": "Environment Config","icon": "⚙", "desc": "Python / framework / git version fingerprint per run.", "status": NEW},
{"id": "llm_log", "label": "LLM Interactions", "icon": "💬", "desc": "Full prompt/response log per step with latency and token counts.", "status": NEW},
{"id": "ml_export", "label": "ML Export", "icon": "⬡", "desc": "Export quality-scored, filtered datasets for model training pipelines.","status": NEW},
],
},
"SETTINGS": {
"accent": "#475569", "icon": "⚙",
"who": "All users",
"description": "Theme, database connection, and host configuration.",
"pages": [
{"id": "settings", "label": "Settings", "icon": "⚙", "desc": "Theme toggle, database path, Live Lab connection, host config.", "status": EXISTS},
],
},
}
# ── Derived lookups (computed once at import) ─────────────────────────────────
# Ordered list of section names
SECTIONS: list[str] = list(SECTION_PAGES.keys())
# page_id → section name
PAGE_TO_SECTION: dict[str, str] = {
page["id"]: section
for section, data in SECTION_PAGES.items()
for page in data["pages"]
}
# page_id → full page metadata dict
PAGE_META: dict[str, dict] = {
page["id"]: {**page, "section": section}
for section, data in SECTION_PAGES.items()
for page in data["pages"]
}
# Set of page_ids with real modules in gui/pages/
PAGES_EXISTING: set[str] = {
pid for pid, meta in PAGE_META.items()
if meta["status"] == EXISTS
}
# Blocked page reasons
PAGES_BLOCKED: dict[str, str] = {
pid: meta["blocked_reason"]
for pid, meta in PAGE_META.items()
if meta.get("blocked_reason")
}
# ── Backward-compat NAV_GROUPS ────────────────────────────────────────────────
# Any code still importing NAV_GROUPS will not break.
NAV_GROUPS = [
item
for section, data in SECTION_PAGES.items()
for item in [(section, None)] + [(p["label"], p["id"]) for p in data["pages"]]
]