Spaces:
Running
Running
File size: 28,683 Bytes
11dde75 ed28626 11dde75 ed28626 11dde75 406b4d1 68b949f 11dde75 ed28626 11dde75 ed28626 b108542 ed28626 6bb7525 ed28626 6bb7525 ed28626 b108542 ed28626 b108542 ed28626 634726e b108542 ed28626 b108542 ed28626 68b949f 67c9044 2cf2ccd 67c9044 2cf2ccd 67c9044 2cf2ccd 67c9044 de5014e 67c9044 de5014e 67c9044 de5014e 830111f de5014e ed28626 6bb7525 ed28626 634726e b108542 68b949f de5014e 67c9044 2cf2ccd de5014e 830111f ed28626 11dde75 ed28626 11dde75 ed28626 11dde75 ed28626 11dde75 ed28626 11dde75 ed28626 | 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 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 | """Run Control — autonomy switch, caps, query frontier. Admin-gated."""
import pandas as pd
import streamlit as st
from agent import agdb, scheduler
from agent import config as C
from agent import ui_theme as T
from agent.ui_common import admin_gate, status_banner, df_or_empty
_TITLE = "Run Control"
_SUB = "Autonomy switch, per-cycle budget, discovery sources and the query frontier."
# Page-local CSS: card_title() applies uppercase/letter-spacing to the whole
# title row; pills/chips passed as right_html must keep their own casing.
# (candidate to hoist into ui_theme.card_title)
st.html("""<style>
.aim-card-title .aim-rc-right { text-transform: none; letter-spacing: 0; display: inline-flex;
gap: 6px; align-items: center; flex-wrap: wrap; justify-content: flex-end; }
.aim-card-title .aim-rc-right .aim-chips { margin: 0; }
.aim-rc-note { font-size: 12.5px; color: var(--aim-ink2); background: var(--aim-tint);
border: 1px solid var(--aim-hairline); border-radius: 8px; padding: 6px 10px; margin: -.3rem 0 .6rem 0;
display: inline-flex; gap: 6px; align-items: center; }
</style>""")
def _right(html: str) -> str:
"""Wrap pill/chip HTML for card_title(right_html=...) so it keeps its casing."""
return f'<span class="aim-rc-right">{html}</span>' if html else ""
def _num_sum(df: pd.DataFrame, col: str) -> int:
"""NaN/NULL-proof integer column sum (0 when the column is missing)."""
if col not in df:
return 0
try:
return int(pd.to_numeric(df[col], errors="coerce").fillna(0).sum())
except Exception:
return 0
# Provisional header (no pills yet) so a DB-unreachable error still shows the
# page title; it is replaced with the pill version once the state is known.
_hdr = st.empty()
with _hdr:
T.page_header(_TITLE, _SUB)
status_banner()
admin = admin_gate()
conn = agdb.connect()
try:
cfg = agdb.get_config(conn)
lock = agdb.lock_is_active(conn) # fresh lock only — stale ≠ running
dlock = agdb.lock_is_active(conn, key="discovery_lock")
try:
qstats = agdb.queue_stats(conn)
except Exception:
qstats = {"by_status": {}, "queued_by_source": {}, "last_discovery": None, "oldest_queued": None}
finally:
conn.close()
_autonomy_on = bool(cfg.get("autonomy"))
_interval_h = float(cfg.get("interval_hours", 6))
_nxt = scheduler.next_run_time()
_nxt_txt = _nxt.strftime("%H:%M UTC") if _nxt else "—"
_autonomy_pill = (T.pill_html("ok", f"autonomy ON · every {_interval_h:g} h")
if _autonomy_on else T.pill_html("off", "autonomy OFF"))
_state_pill = (T.pill_html("running", "cycle running") if lock
else T.pill_html("idle", "idle"))
_pills = _autonomy_pill + _state_pill
if not admin:
_pills += T.pill_html("info", "read-only")
with _hdr:
T.page_header(_TITLE, _SUB, right=_pills)
if not admin:
st.html('<div class="aim-rc-note"><span aria-hidden="true">🔒</span>'
'Unlock admin in the sidebar to change settings — everything below is read-only.</div>')
# ---------------------------------------------------------------------------
# Card grid — Row 1: Autonomy | Sources & behavior; Row 2: Per-cycle budget |
# Manual run; then the Save row and the Query frontier. The column containers
# are created up front and filled later so the widget/button order in code
# stays exactly as before: autonomy → budget → sources → save → manual run →
# query frontier.
# ---------------------------------------------------------------------------
row1_left, row1_right = st.columns([5, 3], gap="medium")
row2_left, row2_right = st.columns([5, 3], gap="medium")
# --- autonomy ---------------------------------------------------------------
with row1_left:
with st.container(border=True, height="stretch"):
T.card_title("Autonomy", right_html=_right(_autonomy_pill))
col1, col2 = st.columns([1, 1], vertical_alignment="bottom")
with col1:
autonomy = st.toggle("Autonomous crawling", value=bool(cfg.get("autonomy")),
disabled=not admin, key="rc_autonomy")
with col2:
interval = st.number_input("Interval (hours)", min_value=0.5, max_value=168.0,
value=float(cfg.get("interval_hours", 6)),
step=0.5, disabled=not admin, key="rc_interval")
st.caption("While ON, the agent runs a full discover→dedupe(DOI)→download→"
"extract→publish cycle at this interval, viewers or not. "
"Restarts resume automatically.")
st.html(T.chips_html([
("next scheduled cycle", _nxt_txt, "muted"),
("interval", f"{_interval_h:g} h", "muted"),
]))
# --- per-cycle budget -------------------------------------------------------
with row2_left:
with st.container(border=True, height="stretch"):
T.card_title("Per-cycle budget", sub=f"hard cap {C.HARD_MAX_PDFS_PER_CYCLE} PDFs / cycle")
b1, b2, b3 = st.columns(3, vertical_alignment="bottom")
with b1:
max_pdfs = st.number_input("Max new PDFs per cycle", 1, 25,
int(cfg.get("max_new_pdfs_per_cycle", 5)),
disabled=not admin, key="rc_max_pdfs")
with b2:
per_query = st.number_input("Max results per query/source", 1, 50,
int(cfg.get("max_per_query", 12)),
disabled=not admin, key="rc_per_query",
help="Relevant candidates taken per query from each "
"source. OpenAlex/S2 calls cost the same at 200 "
"rows as at 20, so this is a download-effort knob, "
"not an API-cost knob.")
with b3:
n_queries = st.number_input("Queries per cycle", 1, 10,
int(cfg.get("queries_per_cycle", 3)),
disabled=not admin, key="rc_n_queries")
b4, b5, b6 = st.columns(3, vertical_alignment="bottom")
with b4:
target_pdfs = st.number_input("Target new PDFs per cycle", 1, 25,
int(cfg.get("min_new_pdfs_per_cycle", 10)),
disabled=not admin, key="rc_target_pdfs",
help="Below this the cycle keeps taking the next "
"frontier slice (up to Max rounds); at the cap "
"above it stops regardless. Each extra round is "
"one more search per query per lane.")
with b5:
max_rounds = st.number_input("Max rounds per cycle", 1, 10,
int(cfg.get("max_rounds_per_cycle", 5)),
disabled=not admin, key="rc_max_rounds")
with b6:
workers = st.number_input("Ingest workers", 1, C.HARD_MAX_INGEST_WORKERS,
int(cfg.get("ingest_workers", 3)),
disabled=not admin, key="rc_workers",
help="PDFs extracted concurrently (Gemini calls are "
"latency-bound). Each worker holds one Postgres "
"connection; 1 = one PDF at a time.")
st.caption("Each cycle picks this many queries, asks every enabled source for at most "
"this many results each, downloads at most this many new PDFs, and "
"extracts them with this many parallel workers.")
# --- sources & behavior -----------------------------------------------------
with row1_right:
with st.container(border=True, height="stretch"):
T.card_title("Sources & behavior")
use_oa = st.checkbox("OpenAlex + Unpaywall", value=bool(cfg.get("use_openalex", True)),
disabled=not admin, key="rc_use_openalex")
use_ax = st.checkbox("arXiv", value=bool(cfg.get("use_arxiv", True)),
disabled=not admin, key="rc_use_arxiv")
use_s2 = st.checkbox("Semantic Scholar (bulk, OA-PDF only)",
value=bool(cfg.get("use_semantic_scholar", False)),
disabled=not admin,
help="One bulk call per query returns up to 1,000 papers that "
"have an open-access PDF. Works without a key on a shared "
"rate bucket; an S2_API_KEY secret gives a dedicated quota.",
key="rc_use_s2")
use_ntrs = st.checkbox("NASA NTRS", value=bool(cfg.get("use_ntrs", True)),
disabled=not admin,
help="NASA Technical Reports Server: free JSON API, no key, "
"no rate limit, direct PDF links (Langley thermoplastic "
"composite reports). Slide decks are skipped.",
key="rc_use_ntrs")
use_topics = st.checkbox("OpenAlex topic walk",
value=bool(cfg.get("use_openalex_topics", True)),
disabled=not admin,
help="Each cycle also takes the next page of open-access works "
"in the configured OpenAlex topics (a 1-credit filter call "
"per topic vs 10 credits per full-text search).",
key="rc_use_topics")
use_ds = st.checkbox("Manufacturer datasheets",
value=bool(cfg.get("use_datasheets", True)),
disabled=not admin,
help="Crawl the datasheet seed pages (Victrex, Hexcel, Toray, RTP, …) "
"for new PDFs; each seed is revisited at most once a day.",
key="rc_use_datasheets")
expand = st.checkbox("Gemini query expansion", value=bool(cfg.get("expand_queries", False)),
disabled=not admin,
help="When a cycle finds nothing new, ask Gemini for fresh search intents.",
key="rc_expand")
st.caption("Discovery sources for each cycle; query expansion only kicks in "
"when a cycle yields nothing new. A source that answers 'budget "
"exhausted' is skipped until its reset time (see Runs & Logs).")
# --- candidate queue ---------------------------------------------------------
_qs = qstats.get("by_status", {})
_q_depth = int(_qs.get("queued", 0))
_q_pill = (T.pill_html("running", "discovery running") if dlock
else T.pill_html("ok", f"{_q_depth:,} queued") if _q_depth
else T.pill_html("warn", "queue empty"))
with st.container(border=True):
T.card_title("Candidate queue",
sub="a discovery pass fills it from every lane; cycles drain it",
right_html=_right(_q_pill))
qc1, qc2, qc3, qc4 = st.columns([1.2, 1, 1, 1], vertical_alignment="bottom")
with qc1:
use_queue = st.toggle("Use the candidate queue",
value=bool(cfg.get("use_candidate_queue", True)),
disabled=not admin, key="rc_use_queue",
help="ON: cycles pop queued candidates and only search inline "
"when the queue is empty; a discovery pass runs on the "
"interval below and whenever the queue falls under the "
"low-water mark. OFF: every cycle searches inline (pre-28 Sep).")
with qc2:
disc_hours = st.number_input("Discovery every (hours)", min_value=1.0, max_value=168.0,
value=float(cfg.get("discovery_interval_hours", 24)),
step=1.0, disabled=not admin, key="rc_disc_hours")
with qc3:
low_water = st.number_input("Low-water mark", 0, 5000,
int(cfg.get("queue_low_water", 150)),
disabled=not admin, key="rc_low_water",
help="A cycle that finds fewer queued candidates than this "
"asks for a discovery pass right after it.")
with qc4:
disc_queries = st.number_input("Queries per pass", 1, 500,
int(cfg.get("discovery_queries", 40)),
disabled=not admin, key="rc_disc_queries",
help="Frontier queries searched on every lane per pass; "
"the first N (below) also hit OpenAlex's 10-credit "
"phrase search.")
qd1, qd2, qd3 = st.columns(3, vertical_alignment="bottom")
with qd1:
disc_oa = st.number_input("…of which on OpenAlex phrase search", 0, 500,
int(cfg.get("discovery_openalex_queries", 30)),
disabled=not admin, key="rc_disc_oa")
with qd2:
disc_pages = st.number_input("Topic pages per topic per pass", 0, 200,
int(cfg.get("discovery_topic_pages", 10)),
disabled=not admin, key="rc_disc_pages",
help="OpenAlex topic walk: 50 works per page, 1 credit each.")
with qd3:
disc_per_query = st.number_input("Candidates per query per lane", 1, 200,
int(cfg.get("discovery_per_query", 40)),
disabled=not admin, key="rc_disc_per_query")
_last = qstats.get("last_discovery")
_nxt_d = scheduler.next_discovery_time()
_by_src = qstats.get("queued_by_source", {})
st.html(T.chips_html(
[("queued", f"{_q_depth:,}", "good" if _q_depth else "warn"),
("ingested", f"{int(_qs.get('ingested', 0)):,}", "muted"),
("downloaded", f"{int(_qs.get('downloaded', 0)):,}", "muted"),
("failed", f"{int(_qs.get('failed', 0)):,}", "error" if _qs.get("failed") else "muted"),
("seen", f"{int(_qs.get('seen', 0)):,}", "muted"),
("last pass", T.fmt_ts(_last) or "never", "muted"),
("next pass", _nxt_d.strftime("%b %d %H:%M UTC") if _nxt_d else "—", "muted")]
+ [(f"queued · {k}", f"{v:,}", "muted") for k, v in list(_by_src.items())[:8]]))
if dlock:
st.info("A discovery pass is running right now — follow it in Runs & Logs.")
if st.button("🔍 Run a discovery pass now", disabled=not admin or bool(dlock),
key="rc_disc_now"):
scheduler.run_discovery_now()
st.toast("Discovery pass started in the background.", icon="🔍")
st.success("Discovery pass started — it appears in Runs & Logs as a 'discovery' run.")
st.caption("UDSpace (UD theses) is on by default; CORE needs the free CORE_API_KEY secret "
"and is inactive without it. Semantic Scholar, arXiv, NTRS and OpenAlex follow "
"the Sources & behavior toggles.")
# --- Gemini cost ----------------------------------------------------------------
_THINK_OPTS = ["low", "minimal", "off", "dynamic", "high"]
_think_now = str(cfg.get("gemini_thinking", "low") or "dynamic").lower()
if _think_now not in _THINK_OPTS:
_THINK_OPTS.append(_think_now) # an explicit token budget set by env
_MODEL_OPTS = list(C.MODEL_PRICES)
_model_now = str(cfg.get("gemini_model") or "gemini-3.5-flash").strip()
if _model_now not in _MODEL_OPTS:
_MODEL_OPTS.insert(0, _model_now)
_VMODEL_OPTS = ["(same as text model)"] + list(C.MODEL_PRICES)
_vmodel_now = str(cfg.get("gemini_vision_model") or "").strip() or "(same as text model)"
if _vmodel_now not in _VMODEL_OPTS:
_VMODEL_OPTS.append(_vmodel_now)
def _price_label(m: str) -> str:
pin, pout = C.model_price(m)
return f"{m} (${pin:g} / ${pout:g} per M)"
with st.container(border=True):
T.card_title("Gemini cost", sub="thinking tokens are billed as output (6× the input price "
"on gemini-3.5-flash)")
k1, k2, k3 = st.columns(3, vertical_alignment="bottom")
with k1:
gemini_thinking = st.selectbox("Thinking per call", _THINK_OPTS,
index=_THINK_OPTS.index(_think_now),
disabled=not admin, key="rc_thinking",
help="dynamic = the API default, unbounded (the 28 Sep "
"cycles: ~30k output tokens per PDF for a ~4k-token "
"extraction). low / minimal are Gemini 3 thinking "
"levels; off = budget 0. Applied to every Gemini call "
"(text extraction, figure classification, mining). "
"If the model rejects the option the call is repeated "
"without it.")
with k2:
gemini_model = st.selectbox("Text extraction model", _MODEL_OPTS,
index=_MODEL_OPTS.index(_model_now),
format_func=_price_label,
disabled=not admin, key="rc_model",
help="Every row is stamped with this model. Change it only "
"after comparing on the gold set (cost_ab.py) — the "
"paper's numbers are per model.")
with k3:
gemini_vision_model = st.selectbox("Figure (vision) model", _VMODEL_OPTS,
index=_VMODEL_OPTS.index(_vmodel_now),
format_func=lambda m: m if m.startswith("(")
else _price_label(m),
disabled=not admin, key="rc_vmodel",
help="The one classify call per PDF (and mining, if "
"on). An easy task: the Lite model is ~5x cheaper.")
dedupe_sim = st.slider("Skip near-duplicate PDFs (text similarity ≥)", 0.0, 1.0,
float(cfg.get("dedupe_text_similarity", 0.9) or 0.0), 0.01,
disabled=not admin, key="rc_dedupe",
help="Before any Gemini call, a PDF whose text is this similar to one "
"already ingested (regional datasheet variants, preprint vs "
"published version) is skipped. 0 = off.")
st.caption("Real cost of the 28–29 Sep cycles on gemini-3.5-flash with dynamic thinking and "
"mining on: ~$0.41 per PDF (36k in / 40k out tokens). Watch Cost and $/PDF in "
"Runs & Logs after a change; rows per paper and the flagged share tell you if "
"recall moved.")
# --- figure mining ----------------------------------------------------------
with st.container(border=True):
T.card_title("Figure & graph mining",
sub=f"hard cap {C.HARD_MAX_FIGURES_PER_PDF} figures / PDF",
right_html=_right(T.pill_html("ok", "quarantined by design")))
g1, g2, g3 = st.columns(3, vertical_alignment="bottom")
with g1:
figures_on = st.toggle("Harvest figures each cycle",
value=bool(cfg.get("figures_enabled", True)),
disabled=not admin, key="rc_figures_on")
with g2:
figure_mine = st.toggle("Mine plot/table values",
value=bool(cfg.get("figure_mining", False)),
disabled=not admin, key="rc_figure_mine",
help="OFF (default since 28 Sep) = harvest + classify only: one "
"vision call per PDF, figure linking still works. ON also "
"reads values off property plots and table images — one "
"thinking vision call per figure, into rows that stay "
"quarantined until a human promotes them.")
with g3:
max_figs = st.number_input("Max figures per PDF", 1,
C.HARD_MAX_FIGURES_PER_PDF,
int(cfg.get("max_figures_per_pdf", 12)),
disabled=not admin, key="rc_max_figs")
st.caption("Each ingested PDF also passes through figures.py: plots and table "
"images are detected, captioned and classified (one vision call per "
"PDF), and mineable ones are read for values. A number read off a "
"graph is an estimate — figure rows land as status "
"'figure_estimate' and only a human promote in the Review Queue "
"can publish them.")
link_on = st.toggle("Link text rows to the figures they cite",
value=bool(cfg.get("link_figures", True)),
disabled=not admin, key="rc_link_figures",
help="figure_links.py: a row whose evidence says 'see Fig. 3' "
"gets that figure's id, the citation, a tier score, and "
"the PNG embedded in its image column. Local work, no "
"model call; the row's value and status are untouched.")
with st.container(horizontal=True, vertical_alignment="center", gap="medium"):
if admin and st.button("💾 Save settings", type="primary", key="rc_save"):
conn = agdb.connect()
try:
agdb.set_config(conn, {
"autonomy": autonomy, "interval_hours": float(interval),
"max_new_pdfs_per_cycle": int(max_pdfs),
"max_per_query": int(per_query), "queries_per_cycle": int(n_queries),
"min_new_pdfs_per_cycle": int(target_pdfs),
"max_rounds_per_cycle": int(max_rounds), "ingest_workers": int(workers),
"use_openalex": use_oa, "use_arxiv": use_ax,
"use_semantic_scholar": use_s2, "use_ntrs": use_ntrs,
"expand_queries": expand,
"use_openalex_topics": use_topics, "use_datasheets": use_ds,
"use_candidate_queue": use_queue, "discovery_interval_hours": float(disc_hours),
"queue_low_water": int(low_water), "discovery_queries": int(disc_queries),
"discovery_openalex_queries": int(disc_oa), "discovery_topic_pages": int(disc_pages),
"discovery_per_query": int(disc_per_query),
"figures_enabled": figures_on, "figure_mining": figure_mine,
"gemini_thinking": str(gemini_thinking),
"gemini_model": str(gemini_model),
"gemini_vision_model": ("" if str(gemini_vision_model).startswith("(")
else str(gemini_vision_model)),
"dedupe_text_similarity": float(dedupe_sim),
"max_figures_per_pdf": int(max_figs),
"link_figures": link_on,
})
finally:
conn.close()
scheduler.sync_from_config()
st.toast("Saved — schedule updated.", icon="💾")
st.success("Saved — schedule updated.")
st.rerun()
if admin:
st.caption("Saves autonomy, budget and sources together; the schedule is re-synced immediately.")
# --- manual cycle -----------------------------------------------------------
with row2_right:
with st.container(border=True, height="stretch"):
T.card_title("Manual run", right_html=_right(_state_pill))
if lock:
st.info("A cycle is running right now — watch Runs & Logs.")
if st.button("▶️ Run one cycle now", type="primary",
disabled=not admin or bool(lock), key="rc_run_now"):
scheduler.run_now()
st.toast("Cycle started in the background.", icon="▶️")
st.success("Cycle started in the background — follow it in Runs & Logs.")
st.caption("Fires one full cycle immediately in the background, independent of "
"the autonomy switch. Only one cycle runs at a time; counters are "
"written when it finishes.")
# --- query frontier ---------------------------------------------------------
qdf = df_or_empty(
"SELECT id, enabled, query, origin, times_used, pdfs_found, rows_yielded, "
"last_used_at FROM agent_queries ORDER BY enabled DESC, id")
if qdf.empty:
_q_sub = ""
_q_chips = ""
else:
_n_q = int(len(qdf))
_n_on = int(qdf["enabled"].fillna(False).astype(bool).sum()) if "enabled" in qdf else 0
_n_manual = int((qdf["origin"].astype(str) == "manual").sum()) if "origin" in qdf else 0
_pdfs = _num_sum(qdf, "pdfs_found")
_rows = _num_sum(qdf, "rows_yielded")
_q_sub = f"{_n_q:,} quer{'y' if _n_q == 1 else 'ies'} · {_n_on:,} enabled"
_q_chips = T.chips_html([
("manual", f"{_n_manual:,}", "muted"),
("PDFs found", f"{_pdfs:,}", "muted"),
("rows yielded", f"{_rows:,}", "muted"),
])
with st.container(border=True):
T.card_title("Query frontier", sub=_q_sub, right_html=_right(_q_chips))
if qdf.empty:
T.empty_state("🧭", "No queries yet",
"The frontier is seeded on first boot; add your own search intent below.")
else:
edited = st.data_editor(
qdf, hide_index=True, width='stretch',
disabled=[] if admin else list(qdf.columns),
column_config={
"id": st.column_config.NumberColumn("ID", width=50, format="%d"),
"enabled": st.column_config.CheckboxColumn("On", width=58),
"query": st.column_config.TextColumn("Query"),
"origin": st.column_config.TextColumn("Origin", width=85),
"times_used": st.column_config.NumberColumn("Times used", width=105, format="%d"),
"pdfs_found": st.column_config.NumberColumn("PDFs found", width=105, format="%d"),
"rows_yielded": st.column_config.NumberColumn("Rows yielded", width=115, format="%d"),
"last_used_at": st.column_config.DatetimeColumn(
"Last used", format="MMM D, HH:mm", width=110),
},
key="query_editor")
st.caption("Each cycle takes the least-recently-used enabled queries; edit the text or "
"toggle a query, then save. Only 'On' and 'Query' are stored.")
if admin and st.button("💾 Save query changes", key="rc_save_queries"):
conn = agdb.connect()
try:
with conn.cursor() as cur:
for _, row in edited.iterrows():
cur.execute(
"UPDATE agent_queries SET enabled = %s, query = %s WHERE id = %s",
(bool(row["enabled"]), str(row["query"]), int(row["id"])))
conn.commit()
finally:
conn.close()
st.toast("Queries updated.", icon="💾")
st.success("Queries updated.")
st.rerun()
with st.container(horizontal=True, vertical_alignment="bottom", gap="small"):
new_q = st.text_input("Add a query", placeholder="e.g. LCP glass fiber dielectric properties",
disabled=not admin, key="rc_new_query")
if admin and st.button("➕ Add query", key="rc_add_query") and new_q.strip():
conn = agdb.connect()
try:
agdb.seed_queries(conn, [new_q.strip()], origin="manual")
finally:
conn.close()
st.toast("Query added to the frontier.", icon="➕")
st.rerun()
|