add Eurostat + UK ONS sources; valuation/bubble/economy/bonds groups; aggregate read; market-open header
Three new data sources hooked into the existing SOURCES registry. All
open APIs, no keys:
- EUROSTAT: prefix EUROSTAT:dataset?dim=val&... — current EU bond
yields (Bund/OAT/BTP/EZ) and Eurozone economic indicators that
FRED's OECD-mirror series stopped updating in 2022-2023.
- ONS: prefix ONS:topic/cdid/dataset — current UK CPI, unemployment,
GDP, industrial production. Replaces the 5+ month-stale FRED
LRHUTTTTGBM156S mirror.
New indicator groups in default.toml feed the strategic/fundamental
lens we converged on: valuation (CAPE/Buffett anchors), bubble_watch
(SKEW/VVIX/RSP vs SPY/HYG vs TLT/IPO/crypto), economy (multi-region,
ALL current-or-stale-flagged), bonds (UK/EU/US/JPN sovereign yields).
Indicator panel now opens with an AI "read" interpretation per group
(generated hourly at :07 UTC alongside an aggregate cross-group read
shown in the dashboard header). The aggregate is grounded by a markets
strip — NYSE/LSE/Frankfurt/Tokyo/HK/Shanghai with open/closed LEDs and
next-open countdown, computed locally from each exchange's tz.
Other UX bits: indicator-row tooltips populated from TOML notes;
rows whose last observation is >90 days old get a 'stale' chip;
ghost symbols (in DB but no longer in TOML) filtered out of the
panel; Eurostat/ONS symbols display as short codes rather than the
full API path.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
parent
a10409c02b
commit
1edf9cad41
15 changed files with 1156 additions and 10 deletions
|
|
@ -16,6 +16,8 @@ from app.config import get_settings
|
|||
|
||||
YAHOO_CHART = "https://query1.finance.yahoo.com/v8/finance/chart/{symbol}"
|
||||
FRED_API = "https://api.stlouisfed.org/fred/series/observations"
|
||||
EUROSTAT_API = "https://ec.europa.eu/eurostat/api/dissemination/statistics/1.0/data/{dataset}"
|
||||
ONS_API = "https://www.ons.gov.uk/{topic}/timeseries/{cdid}/{dataset}/data"
|
||||
UA = {"User-Agent": "Mozilla/5.0 (cassandra) Python/httpx"}
|
||||
|
||||
|
||||
|
|
@ -212,10 +214,225 @@ async def fetch_fred(
|
|||
return Quote(symbol, "fred", label, note, None, None, None, error=str(e))
|
||||
|
||||
|
||||
# --- Eurostat (no API key needed) -------------------------------------------
|
||||
|
||||
|
||||
def _eurostat_time_to_iso(t: str) -> str:
|
||||
"""Convert Eurostat time codes into ISO-style dates so they sort and
|
||||
compare correctly. Accepts YYYY-MM, YYYY-Qn, YYYY, and YYYY-MM-DD."""
|
||||
t = t.strip()
|
||||
if len(t) == 4 and t.isdigit(): # annual: "2026"
|
||||
return f"{t}-01-01"
|
||||
if len(t) == 6 and t[4] == "Q": # quarterly: "2026Q1"
|
||||
q = int(t[5])
|
||||
return f"{t[:4]}-{(q - 1) * 3 + 1:02d}-01"
|
||||
if len(t) == 7 and t[4] == "-": # monthly: "2026-03"
|
||||
return f"{t}-01"
|
||||
if len(t) == 10: # daily: "2026-03-15"
|
||||
return t
|
||||
return t # fall through; caller may flag
|
||||
|
||||
|
||||
async def fetch_eurostat(
|
||||
client: httpx.AsyncClient,
|
||||
symbol: str,
|
||||
label: str,
|
||||
note: str,
|
||||
anchor: str | None = None,
|
||||
) -> Quote:
|
||||
"""Fetch a Eurostat time series. `symbol` format:
|
||||
DATASET?dim1=val1&dim2=val2
|
||||
e.g. 'irt_lt_mcby_m?geo=DE&int_rt=MCBY' for German 10y bond yield.
|
||||
Eurostat's API is open (no key), uses JSON-stat 2.0."""
|
||||
import urllib.parse
|
||||
|
||||
try:
|
||||
if "?" in symbol:
|
||||
dataset, query = symbol.split("?", 1)
|
||||
params = dict(urllib.parse.parse_qsl(query))
|
||||
else:
|
||||
dataset, params = symbol, {}
|
||||
params.setdefault("format", "JSON")
|
||||
params.setdefault("lang", "EN")
|
||||
|
||||
r = await client.get(
|
||||
EUROSTAT_API.format(dataset=dataset),
|
||||
params=params, headers=UA, timeout=20,
|
||||
)
|
||||
r.raise_for_status()
|
||||
data = r.json()
|
||||
|
||||
time_cat = data["dimension"]["time"]["category"]
|
||||
# JSON-stat 2.0: {"index": {timecode: pos}, "label": {timecode: human}}
|
||||
time_index = time_cat["index"]
|
||||
values = data.get("value") or {}
|
||||
|
||||
# Build (iso_date, value) pairs, sorted ascending in time.
|
||||
rows: list[tuple[str, float]] = []
|
||||
for tcode, pos in sorted(time_index.items(), key=lambda kv: kv[1]):
|
||||
raw = values.get(str(pos))
|
||||
if raw is None:
|
||||
continue
|
||||
try:
|
||||
rows.append((_eurostat_time_to_iso(tcode), float(raw)))
|
||||
except (TypeError, ValueError):
|
||||
continue
|
||||
|
||||
if not rows:
|
||||
raise ValueError("no observations")
|
||||
|
||||
last_date, last_val = rows[-1]
|
||||
|
||||
def _find_back(min_days: int) -> float | None:
|
||||
ref = datetime.strptime(last_date, "%Y-%m-%d").date()
|
||||
for d, v in reversed(rows[:-1]):
|
||||
if (ref - datetime.strptime(d, "%Y-%m-%d").date()).days >= min_days:
|
||||
return v
|
||||
return None
|
||||
|
||||
prev_val = rows[-2][1] if len(rows) >= 2 else None
|
||||
changes = {
|
||||
"1d": _pct(prev_val, last_val),
|
||||
"1m": _pct(_find_back(28), last_val),
|
||||
"1y": _pct(_find_back(360), last_val),
|
||||
}
|
||||
anchor_used: str | None = None
|
||||
if anchor:
|
||||
anchor_d = _parse_date(anchor).date()
|
||||
for d, v in reversed(rows):
|
||||
if datetime.strptime(d, "%Y-%m-%d").date() <= anchor_d:
|
||||
changes["anchor"] = _pct(v, last_val)
|
||||
anchor_used = d
|
||||
break
|
||||
|
||||
return Quote(
|
||||
symbol=symbol, source="eurostat", label=label, note=note,
|
||||
price=last_val, currency=None, as_of=last_date,
|
||||
changes=changes, anchor_date=anchor_used,
|
||||
)
|
||||
except Exception as e:
|
||||
return Quote(symbol, "eurostat", label, note, None, None, None, error=str(e))
|
||||
|
||||
|
||||
# --- UK ONS (Office for National Statistics, no API key needed) -------------
|
||||
|
||||
|
||||
_ONS_MONTH = {
|
||||
"JAN": 1, "FEB": 2, "MAR": 3, "APR": 4, "MAY": 5, "JUN": 6,
|
||||
"JUL": 7, "AUG": 8, "SEP": 9, "OCT": 10, "NOV": 11, "DEC": 12,
|
||||
}
|
||||
|
||||
|
||||
def _ons_date_to_iso(s: str) -> str | None:
|
||||
"""ONS date formats: monthly '2026 MAR', quarterly '2026 Q1', annual '2025'."""
|
||||
s = s.strip().upper()
|
||||
parts = s.split()
|
||||
try:
|
||||
if len(parts) == 1 and parts[0].isdigit():
|
||||
return f"{parts[0]}-01-01"
|
||||
if len(parts) == 2:
|
||||
year = int(parts[0])
|
||||
tag = parts[1]
|
||||
if tag in _ONS_MONTH:
|
||||
return f"{year:04d}-{_ONS_MONTH[tag]:02d}-01"
|
||||
if tag.startswith("Q") and tag[1:].isdigit():
|
||||
q = int(tag[1:])
|
||||
return f"{year:04d}-{(q - 1) * 3 + 1:02d}-01"
|
||||
except (ValueError, IndexError):
|
||||
pass
|
||||
return None
|
||||
|
||||
|
||||
async def fetch_ons(
|
||||
client: httpx.AsyncClient,
|
||||
symbol: str,
|
||||
label: str,
|
||||
note: str,
|
||||
anchor: str | None = None,
|
||||
) -> Quote:
|
||||
"""Fetch a UK ONS time series. `symbol` format:
|
||||
<topic_path>/<cdid>/<dataset>
|
||||
e.g. 'economy/inflationandpriceindices/d7g7/mm23' for UK CPI YoY.
|
||||
ONS publishes via www.ons.gov.uk; no auth, JSON when Accept header set."""
|
||||
try:
|
||||
parts = symbol.split("/")
|
||||
if len(parts) < 3:
|
||||
raise ValueError("ONS symbol must be topic/cdid/dataset")
|
||||
dataset = parts[-1]
|
||||
cdid = parts[-2]
|
||||
topic = "/".join(parts[:-2])
|
||||
|
||||
r = await client.get(
|
||||
ONS_API.format(topic=topic, cdid=cdid, dataset=dataset),
|
||||
headers={**UA, "Accept": "application/json"},
|
||||
timeout=20,
|
||||
)
|
||||
r.raise_for_status()
|
||||
data = r.json()
|
||||
|
||||
# Use the most granular series available: months > quarters > years.
|
||||
for key in ("months", "quarters", "years"):
|
||||
raw_seq = data.get(key) or []
|
||||
if raw_seq:
|
||||
break
|
||||
if not raw_seq:
|
||||
raise ValueError("no observations")
|
||||
|
||||
rows: list[tuple[str, float]] = []
|
||||
for entry in raw_seq:
|
||||
iso = _ons_date_to_iso(entry.get("date", ""))
|
||||
v = entry.get("value")
|
||||
if iso is None or v in (None, "", "."):
|
||||
continue
|
||||
try:
|
||||
rows.append((iso, float(v)))
|
||||
except (TypeError, ValueError):
|
||||
continue
|
||||
if not rows:
|
||||
raise ValueError("no parseable observations")
|
||||
|
||||
last_date, last_val = rows[-1]
|
||||
|
||||
def _find_back(min_days: int) -> float | None:
|
||||
ref = datetime.strptime(last_date, "%Y-%m-%d").date()
|
||||
for d, v in reversed(rows[:-1]):
|
||||
if (ref - datetime.strptime(d, "%Y-%m-%d").date()).days >= min_days:
|
||||
return v
|
||||
return None
|
||||
|
||||
prev_val = rows[-2][1] if len(rows) >= 2 else None
|
||||
changes = {
|
||||
"1d": _pct(prev_val, last_val),
|
||||
"1m": _pct(_find_back(28), last_val),
|
||||
"1y": _pct(_find_back(360), last_val),
|
||||
}
|
||||
anchor_used: str | None = None
|
||||
if anchor:
|
||||
anchor_d = _parse_date(anchor).date()
|
||||
for d, v in reversed(rows):
|
||||
if datetime.strptime(d, "%Y-%m-%d").date() <= anchor_d:
|
||||
changes["anchor"] = _pct(v, last_val)
|
||||
anchor_used = d
|
||||
break
|
||||
|
||||
return Quote(
|
||||
symbol=symbol, source="ons", label=label, note=note,
|
||||
price=last_val, currency=None, as_of=last_date,
|
||||
changes=changes, anchor_date=anchor_used,
|
||||
)
|
||||
except Exception as e:
|
||||
return Quote(symbol, "ons", label, note, None, None, None, error=str(e))
|
||||
|
||||
|
||||
# --- Source registry ----------------------------------------------------------
|
||||
|
||||
FetcherFn = Callable[..., "Quote"]
|
||||
SOURCES: dict[str, FetcherFn] = {"yahoo": fetch_yahoo, "FRED": fetch_fred}
|
||||
SOURCES: dict[str, FetcherFn] = {
|
||||
"yahoo": fetch_yahoo,
|
||||
"FRED": fetch_fred,
|
||||
"EUROSTAT": fetch_eurostat,
|
||||
"ONS": fetch_ons,
|
||||
}
|
||||
|
||||
|
||||
def parse_symbol(symbol: str) -> tuple[FetcherFn, str]:
|
||||
|
|
|
|||
84
app/services/markets.py
Normal file
84
app/services/markets.py
Normal file
|
|
@ -0,0 +1,84 @@
|
|||
"""Market-open/close status for the dashboard header. Pure computation —
|
||||
no API needed; the schedules are known constants. Holidays are NOT modelled
|
||||
(would require a region-specific calendar); a closed Monday will still show
|
||||
"open" if the time-of-day fits. Good enough for the strategic dashboard.
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
from dataclasses import dataclass
|
||||
from datetime import datetime, time, timedelta, timezone
|
||||
from zoneinfo import ZoneInfo
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class Market:
|
||||
code: str
|
||||
name: str
|
||||
tz: str # IANA zone (handles DST automatically)
|
||||
open: time # local time
|
||||
close: time # local time
|
||||
|
||||
|
||||
# Mon=0 .. Sun=6. Markets observe Mon-Fri unless overridden.
|
||||
_WORKWEEK = {0, 1, 2, 3, 4}
|
||||
|
||||
|
||||
MARKETS: list[Market] = [
|
||||
Market("NYSE", "NYSE", "America/New_York", time(9, 30), time(16, 0)),
|
||||
Market("LSE", "LSE", "Europe/London", time(8, 0), time(16, 30)),
|
||||
Market("XETRA", "Frankfurt","Europe/Berlin", time(9, 0), time(17, 30)),
|
||||
Market("JPX", "Tokyo", "Asia/Tokyo", time(9, 0), time(15, 0)),
|
||||
Market("HKEX", "Hong Kong","Asia/Hong_Kong", time(9, 30), time(16, 0)),
|
||||
Market("SSE", "Shanghai", "Asia/Shanghai", time(9, 30), time(15, 0)),
|
||||
]
|
||||
|
||||
|
||||
def _next_open_at(m: Market, now_utc: datetime) -> datetime:
|
||||
"""Earliest future open datetime (UTC) for this market, scanning ahead
|
||||
up to 7 days for the next weekday."""
|
||||
tz = ZoneInfo(m.tz)
|
||||
local = now_utc.astimezone(tz)
|
||||
candidate_date = local.date()
|
||||
for _ in range(8): # today + 7 days
|
||||
weekday = candidate_date.weekday()
|
||||
if weekday in _WORKWEEK:
|
||||
local_open = datetime.combine(candidate_date, m.open, tzinfo=tz)
|
||||
if local_open > local:
|
||||
return local_open.astimezone(timezone.utc)
|
||||
candidate_date = candidate_date + timedelta(days=1)
|
||||
return now_utc + timedelta(days=7) # fallback (shouldn't happen)
|
||||
|
||||
|
||||
def _close_at(m: Market, now_utc: datetime) -> datetime:
|
||||
"""Today's close in UTC (assumes we've already established it's open)."""
|
||||
tz = ZoneInfo(m.tz)
|
||||
local = now_utc.astimezone(tz)
|
||||
return datetime.combine(local.date(), m.close, tzinfo=tz).astimezone(timezone.utc)
|
||||
|
||||
|
||||
def status_for(m: Market, now_utc: datetime) -> dict:
|
||||
tz = ZoneInfo(m.tz)
|
||||
local = now_utc.astimezone(tz)
|
||||
is_workday = local.weekday() in _WORKWEEK
|
||||
in_session = is_workday and m.open <= local.time() < m.close
|
||||
if in_session:
|
||||
return {
|
||||
"code": m.code,
|
||||
"name": m.name,
|
||||
"open": True,
|
||||
"until": _close_at(m, now_utc),
|
||||
"label": "open",
|
||||
}
|
||||
return {
|
||||
"code": m.code,
|
||||
"name": m.name,
|
||||
"open": False,
|
||||
"until": _next_open_at(m, now_utc),
|
||||
"label": "closed",
|
||||
}
|
||||
|
||||
|
||||
def all_statuses(now_utc: datetime | None = None) -> list[dict]:
|
||||
if now_utc is None:
|
||||
now_utc = datetime.now(timezone.utc)
|
||||
return [status_for(m, now_utc) for m in MARKETS]
|
||||
|
|
@ -20,7 +20,7 @@ OPENROUTER_URL = "https://openrouter.ai/api/v1/chat/completions"
|
|||
# Bump when the composed prompt changes meaningfully. Stored on every
|
||||
# StrategicLog row so historical logs can be linked to the prompt that produced
|
||||
# them.
|
||||
PROMPT_VERSION = 3
|
||||
PROMPT_VERSION = 4
|
||||
|
||||
|
||||
# --- Core: invariant across tone/analysis settings ----------------------------
|
||||
|
|
@ -60,6 +60,28 @@ numbers in every paragraph. No section over ~150 words.
|
|||
- End with a watch list: 3-5 specific items to track in the next week, \
|
||||
each one sentence.
|
||||
|
||||
# Time-horizon discipline
|
||||
- This is a STRATEGIC log, not a day-trader's read. Treat 1-day moves under \
|
||||
2% as background noise; mention them only when they break or confirm a \
|
||||
multi-week trend or are extreme outliers.
|
||||
- Anchor every claim to multi-week (1m), multi-month (since-anchor), or \
|
||||
multi-year (1y) changes — not 1d. If the only thing happening is a 1d move, \
|
||||
omit the paragraph.
|
||||
- The watch list is for "structural tripwires over the next 1-3 months", not \
|
||||
"things to watch tomorrow". Each watch item should name a level/threshold \
|
||||
whose breach would change the regime, not a calendar-date event.
|
||||
|
||||
# Rational vs irrational framing
|
||||
The reader's primary goal is to disconnect rational decisions from market \
|
||||
irrationality. In every sector or theme paragraph, separately identify:
|
||||
- The RATIONAL drivers: earnings, real-economy data, monetary policy, \
|
||||
structural geopolitical shifts, valuation vs fundamentals.
|
||||
- The IRRATIONAL drivers: positioning, narrative momentum, sentiment \
|
||||
extremes, concentration, flow-driven moves, options gamma, credit complacency.
|
||||
When the two diverge — price moving on irrational drivers while fundamentals \
|
||||
say otherwise, or vice versa — flag the divergence explicitly. Those gaps \
|
||||
are where the next regime change starts.
|
||||
|
||||
# Discipline
|
||||
- No emojis, no marketing language, no "concerning" or "unprecedented" \
|
||||
without a specific number behind it.
|
||||
|
|
@ -68,7 +90,16 @@ without a specific number behind it.
|
|||
predicted X and X did not happen". Both are useful; conflating them is not.
|
||||
- Don't repeat the same point in different words across paragraphs.
|
||||
- No buy/sell recommendations. Triggers are pre-set elsewhere; your job is \
|
||||
to report whether reality is confirming, modifying, or refuting the thesis."""
|
||||
to report whether reality is confirming, modifying, or refuting the thesis.
|
||||
|
||||
# System temperature (closing line, mandatory)
|
||||
Close the log with a single sentence on a line of its own, formatted exactly:
|
||||
|
||||
System temperature: [cool|neutral|elevated|hot|extreme] — [one clause naming the 2-3 specific divergences or readings that justify the label]
|
||||
|
||||
This is the line a reader who only sees the watch list scrolls down to. Make \
|
||||
it earn its place: cite real signals (HY OAS, breadth, VIX, valuation, real \
|
||||
yields), not vibes."""
|
||||
|
||||
|
||||
# --- Tone: audience-shaping block --------------------------------------------
|
||||
|
|
@ -141,6 +172,118 @@ question via the chat sidebar.
|
|||
- Keep the same audience and analysis discipline established above."""
|
||||
|
||||
|
||||
def build_summary_system_prompt(tone: str, analysis: str) -> str:
|
||||
"""A lean, focused system prompt for the per-indicator-group hourly
|
||||
summary. INTERPRETATION not description — the reader has the table
|
||||
next to this paragraph; they don't need numbers recited at them."""
|
||||
tone_block = _TONE.get(tone.upper(), _TONE["INTERMEDIATE"])
|
||||
analysis_block = _ANALYSIS.get(analysis.upper(), _ANALYSIS["SPECULATIVE"])
|
||||
return f"""You write a TINY interpretation (≤60 words, 2-3 sentences) \
|
||||
of ONE indicator group for a strategic markets dashboard.
|
||||
|
||||
# What this is for
|
||||
The reader is looking at the table of numbers right next to your text. \
|
||||
They can see the values. They CANNOT see the meaning. Your job is to \
|
||||
**explain what the data means**, not to recite it. Each sentence should be \
|
||||
a regime-level interpretation, a fundamental driver identification, or a \
|
||||
cross-indicator implication — not a description of moves.
|
||||
|
||||
# Hard constraints
|
||||
- Plain prose, ONE paragraph. No markdown, no headers, no lists, no labels.
|
||||
- Open IMMEDIATELY with substance. NEVER start with: "I need to", "I'll", \
|
||||
"We need to", "We are asked", "Here's", "Let me", "Let's", "Sure", "Looking \
|
||||
at", "Based on", "Summary:", "The data shows", "First", "To address". No \
|
||||
meta-commentary at all.
|
||||
- Cite at most 2-3 specific numbers and ONLY when they anchor an \
|
||||
interpretation. Don't list moves; explain them.
|
||||
- Multi-week / multi-month horizon. 1-day moves under 2% are noise — skip.
|
||||
- No buy/sell language. No predictions. No watch list. No TL;DR. No date \
|
||||
header. No "system temperature" line — that belongs to the full daily log.
|
||||
|
||||
{tone_block}
|
||||
|
||||
{analysis_block}
|
||||
|
||||
# Bad example — describes what happened
|
||||
"S&P +5.2% 1m and Nasdaq +8.8% 1m diverge from FTSE -3.4% and Euro Stoxx \
|
||||
-2.6%. The US-vs-rest gap is widening."
|
||||
|
||||
# Good example — interprets what it means
|
||||
"The US-vs-rest equity gap is funded by AI-capex concentration in 7 names; \
|
||||
the breadth-weighted RSP barely keeps pace with SPY, which is the classic \
|
||||
late-cycle marker — narrow leadership, not broad recovery. The 5% 1m gap \
|
||||
between Nasdaq and FTSE is a narrative trade, not a fundamental one."
|
||||
"""
|
||||
|
||||
|
||||
def build_summary_user_prompt(group_name: str, quotes: list[dict]) -> str:
|
||||
parts = [
|
||||
f"# Group: {group_name}",
|
||||
"Indicators (latest reading + 1d/1m/1y/since-anchor change):",
|
||||
"```json",
|
||||
json.dumps(quotes, indent=2, default=str)[:12000],
|
||||
"```",
|
||||
"\nWrite the 2-3 sentence read for this group now.",
|
||||
]
|
||||
return "\n".join(parts)
|
||||
|
||||
|
||||
def build_aggregate_summary_system_prompt(tone: str, analysis: str) -> str:
|
||||
"""System prompt for the cross-group aggregate read shown on the dashboard.
|
||||
Wider lens than a per-group summary — synthesise across all groups."""
|
||||
tone_block = _TONE.get(tone.upper(), _TONE["INTERMEDIATE"])
|
||||
analysis_block = _ANALYSIS.get(analysis.upper(), _ANALYSIS["SPECULATIVE"])
|
||||
return f"""You write a single SHORT cross-asset INTERPRETATION (≤80 \
|
||||
words, 2-4 sentences) for the dashboard header. The reader is glancing — \
|
||||
give them the meaning of the whole tape, not a recap.
|
||||
|
||||
# What this is for
|
||||
The reader can see every indicator on the dashboard below this paragraph. \
|
||||
Your job is NOT to summarise the moves. It is to explain what the moves, \
|
||||
**taken together as a system**, mean: which regime is being signalled, \
|
||||
which divergences are load-bearing, what fundamental story the cross-asset \
|
||||
behaviour tells.
|
||||
|
||||
# Hard constraints
|
||||
- Plain prose, ONE paragraph. No markdown, headers, lists, or labels.
|
||||
- Open IMMEDIATELY with substance. NEVER start with: "I need to", "I'll", \
|
||||
"We need to", "Here's", "Let me", "Looking at", "Based on", "Sure", "Summary:", \
|
||||
"The data shows", "Across the board". No meta-commentary.
|
||||
- Identify the single most important **cross-asset implication**: e.g. \
|
||||
"rates and credit disagree", "equities outrun fundamentals", "geopolitical \
|
||||
risk premium is in commodities but not vol". Cite no more than 3 specific \
|
||||
numbers, and only as anchors for the interpretation.
|
||||
- Multi-week / multi-month horizon. 1-day moves under 2% are noise.
|
||||
- No buy/sell language. No predictions of specific levels.
|
||||
|
||||
{tone_block}
|
||||
|
||||
{analysis_block}
|
||||
|
||||
# Bad example — describes
|
||||
"Equities are up, real yields are higher, HY OAS is tight, breadth is \
|
||||
narrowing."
|
||||
|
||||
# Good example — interprets
|
||||
"The tape is paying a rising real discount rate (US 10y real +15bp 1m) with \
|
||||
conviction for AI growth, but credit refuses to confirm and breadth is \
|
||||
narrowing — that combination is what late-cycle looks like, not pre-crash. \
|
||||
The risk is not the level but the convergence: if any one of credit, \
|
||||
breadth, or vol turns, the others will follow fast."
|
||||
"""
|
||||
|
||||
|
||||
def build_aggregate_summary_user_prompt(quotes_by_group: dict[str, list[dict]]) -> str:
|
||||
parts = [
|
||||
"# All indicator groups (latest readings + change windows)",
|
||||
"```json",
|
||||
json.dumps(quotes_by_group, indent=2, default=str)[:20000],
|
||||
"```",
|
||||
"\nWrite the cross-asset aggregate read now.",
|
||||
]
|
||||
return "\n".join(parts)
|
||||
|
||||
|
||||
def build_chat_system_prompt(
|
||||
tone: str,
|
||||
analysis: str,
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue