Feed API
Signals via API
Feed API access for integrations is available under Business plans. Public endpoints show a limited or redacted payload; Pro is for individual app access.
Public sample from the feed endpoint. The backend exposes read endpoints for signals, storylines, and briefings.
Public sample
GET /v1/feed/stories?limit=3&timeframe=24h&sort=momentum&tenant=biotech
{
"run_id": "dc190557-289a-46d1-a3ce-38d6df741df2",
"timeframe": "24h",
"items": [
{
"type": "storyline",
"id": "51c816c8-e17b-4b62-a92d-d4674f888a68",
"story_id": "51c816c8-e17b-4b62-a92d-d4674f888a68",
"run_id": "dc190557-289a-46d1-a3ce-38d6df741df2",
"snapshot_id": "8a599dd4-1f1a-4a5e-96c9-b78d6d2aa80e",
"narrative_id": "df96691c-481a-4ff2-a70d-4b4df4a2cf65",
"first_seen": "2026-10-02T17:05:18.661724+00:00",
"last_seen": "2026-10-03T05:05:23.506621+00:00",
"first_seen_at": "2026-10-02T08:24:00+00:00",
"last_seen_at": "2026-10-02T13:01:00+00:00",
"status": "open",
"title": "Human medicines European public assessment report (EPAR): Byetta, exenatide, Date of authorisation: 20/11/2006, Revision: 30, Status: Withdrawn",
"summary": "Human medicines European public assessment report (EPAR): Byetta, exenatide, Date of authorisation: 20/11/2006, Revision: 30, Status: Withdrawn",
"top_tickers": [],
"why_now": [],
"metrics": {
"score_total": 0.858152,
"momentum_24h": 4,
"post_count": 4,
"unique_origin_publishers": 1,
"unique_publisher_types": 1,
"origin_share_top1": 1,
"amplifier_share_top1": null,
"duplicate_ratio": 0,
"evidence_score": 0.25,
"source_types_count": 1,
"unique_origin_domains": 1,
"unique_publishers": 1,
"quality_policy_version": 6,
"tenant_claim_coherence_policy_version": 3
},
"trend_status": "insufficient_history",
"trend_sparkline": "▁▁",
"trend_points_n": 2,
"trend_window": {
"lookback_days": 14,
"max_points": 36,
"value_key": "score_total"
},
"maturity_label": "seed",
"maturity_score": 0.22918691666666666,
"top_sources": [
{
"label": "EMA (via EMA What's new): Human medicines European public assessment report (EPAR): Ordspono, odronextamab, Date of authorisation: 22/08/2024, Revision: 2, Status: Authorised",
"source_type": "rss",
"evidence_class": "primary",
"provenance_role": "origin",
"publisher_subtype": "regulator"
}
],
"rank_story": 0.697245690203094,
"p_score": 1,
"p_mom": 1,
"recency": 0.5889827608123758,
"risk_penalty": 0.1,
"badges": [
"seed",
"insufficient_history",
"low_evidence"
],
"storyline_category": "narrative",
"storyline_category_reason": "multi_run_continuity",
"verified_update_issuer_count": 1,
"storyline_independent_origin_count": 1,
"storyline_continuity_points": 2,
"sources": [],
"llm_title": "EMA revises EPAR records for four medicines",
"llm_summary": "EMA’s latest EPAR entries report revisions for four medicines. Sialanar, Ordspono and Riltrava Aerosphere are marked authorised, while Byetta is marked withdrawn. The posts provide current record statuses but do not explain the changes.",
"llm_narrative_frame": "The latest EMA EPAR updates show revised records for four medicines. Sialanar, Ordspono and Riltrava Aerosphere remain listed as authorised, while Byetta is listed as withdrawn.",
"llm_narrative_label": "EMA medicine-status updates",
"llm_narrative_type": "ongoing_storyline",
"llm_why_now": [
"Four revised EPAR entries appeared in the latest EMA updates.",
"The entries include revision numbers alongside current statuses."
],
"llm_why_it_matters": [
"The entries provide updated EMA record statuses for four medicines.",
"The batch includes both authorised and withdrawn medicine records."
],
"llm_top_sources": [
{
"label": "Sialanar EPAR"
},
{
"label": "Ordspono EPAR"
},
{
"label": "Riltrava Aerosphere EPAR"
},
{
"label": "Byetta EPAR"
}
],
"entities": {
"projects": [
"Sialanar",
"Ordspono",
"Riltrava Aerosphere",
"Byetta"
]
},
"recurring_claims": [
{
"claim": "EMA records list Sialanar, Ordspono and Riltrava Aerosphere as authorised after revisions.",
"evidence_urls": [
"https://ema.europa.eu/en/medicines/human/EPAR/sialanar",
"https://ema.europa.eu/en/medicines/human/EPAR/ordspono",
"https://ema.europa.eu/en/medicines/human/EPAR/riltrava-aerosphere"
]
},
{
"claim": "The EMA record for Byetta lists the medicine as withdrawn after revision 30.",
"evidence_urls": [
"https://ema.europa.eu/en/medicines/human/EPAR/byetta"
]
}
],
"stance_map": [
{
"who": "EMA",
"stance": "neutral",
"evidence_urls": [
"https://ema.europa.eu/en/medicines/human/EPAR/sialanar",
"https://ema.europa.eu/en/medicines/human/EPAR/ordspono",
"https://ema.europa.eu/en/medicines/human/EPAR/riltrava-aerosphere"
]
},
{
"who": "EMA",
"stance": "neutral",
"evidence_urls": [
"https://ema.europa.eu/en/medicines/human/EPAR/byetta"
]
}
],
"quality_flags": {
"mixed_topic_risk": "low",
"promo_risk": "low",
"source_quality": "high"
},
"editor_note": "Four EMA EPAR records were revised, with three medicines listed as authorised and one as withdrawn.",
"scope_tags": [
"regulatory",
"EMA",
"medicines",
"EPAR",
"authorisation",
"withdrawal"
],
"scope_tags_raw": [
"regulatory",
"EMA",
"medicines",
"EPAR",
"authorisation",
"withdrawal"
],
"llm_status": "accepted",
"llm_meta": {
"model": "gpt-5.6-luna",
"prompt_version": "enrich_v3",
"input_hash": "f8e465f145e016c5cbab1ebaa8f5bd96d2a8d9738c66a02857c40c73dbbaf1e5",
"updated_at": "2026-10-03T05:08:00.476277+00:00",
"llm_metadata": {
"env": "prod",
"host": "83bc069ed062",
"stage": "signals.enrich",
"run_id": "dc190557-289a-46d1-a3ce-38d6df741df2",
"tenant": "biotech",
"service": "api",
"pipeline": "pipeline_run",
"correlation_id": "df96691c-481a-4ff2-a70d-4b4df4a2cf65"
}
},
"display_title": "EMA revises EPAR records for four medicines",
"display_summary": "EMA’s latest EPAR entries report revisions for four medicines. Sialanar, Ordspono and Riltrava Aerosphere are marked authorised, while Byetta is marked withdrawn. The posts provide current record statuses but do not explain the changes.",
"display_tags": [
"regulatory",
"ema",
"medicines",
"epar",
"authorisation",
"withdrawal"
],
"tags": [
"regulatory",
"ema",
"medicines",
"epar",
"authorisation",
"withdrawal"
],
"cscope_tags": [
"regulatory",
"ema",
"medicines",
"epar",
"authorisation",
"withdrawal"
],
"display_why_now": [
"Four revised EPAR entries appeared in the latest EMA updates.",
"The entries include revision numbers alongside current statuses."
],
"display_why_it_matters": [
"The entries provide updated EMA record statuses for four medicines.",
"The batch includes both authorised and withdrawn medicine records."
],
"why_now_display": [
"Four revised EPAR entries appeared in the latest EMA updates.",
"The entries include revision numbers alongside current statuses."
],
"why_it_matters_display": [
"The entries provide updated EMA record statuses for four medicines.",
"The batch includes both authorised and withdrawn medicine records."
],
"show_why": true,
"sources_display": [
{
"label": "ema.europa.eu: Ordspono EPAR"
}
],
"editor_status": "passed",
"fields_replaced": [],
"narrative_frame_display": "The latest EMA EPAR updates show revised records for four medicines. Sialanar, Ordspono and Riltrava Aerosphere remain listed as authorised, while Byetta is listed as withdrawn.",
"display_rank_base_score": 4,
"display_rank_score": 1000000004,
"display_rank_boost": 1000000000,
"lane": "chatter",
"lane_reason": "maturity=seed",
"evidence_mix": {
"primary": 1,
"secondary": 0,
"specialist": 0,
"aggregator": 0,
"social": 0
},
"sourceStats": {
"qualityPolicyVersion": 6,
"evidenceDocumentCount": 4,
"uniqueItemCount": 4,
"uniquePublisherCount": 1,
"uniqueCanonicalOriginCount": 1,
"independentNonSocialCount": 1,
"primaryCount": 1,
"secondaryCount": 0,
"uniqueDomainsCount": 1,
"aggregatorCount": 0,
"socialCount": 0,
"sourceTypeDiversityCount": 1,
"topSourceShare": 1,
"passesTopSignalsGate": false,
"gateReason": "independentNonSocial=1; primary=1; secondary=0; documents=4; publishers=1; canonicalOrigins=1; topSourceShare=1.0; rule=>=2 unique canonical primary/secondary publishers; aggregators/social excluded"
},
"evidence_meta": {
"post_count": 4,
"unique_origin_publishers": 1,
"source_types_count": 1,
"origin_share_top1": 1
},
"gate": {
"passesTopSignalsGate": false,
"gateReason": "independentNonSocial=1; primary=1; secondary=0; documents=4; publishers=1; canonicalOrigins=1; topSourceShare=1.0; rule=>=2 unique canonical primary/secondary publishers; aggregators/social excluded"
},
"tenant_claim_coherence": {
"tenant": "biotech",
"status": "not_applicable",
"claim_scope": "not_applicable",
"reason_codes": [],
"public_anchors": {},
"evidence_anchors": {},
"debug": {
"evidence_record_count": 4,
"matching_record_count": 0,
"independent_support_count": 0,
"conflicting_record_count": 0
}
},
"evidence_quality_flag": "weak"
},
{
"type": "storyline",
"id": "ae1f5d53-0b90-4329-ad19-8e7d1fbe803a",
"story_id": "ae1f5d53-0b90-4329-ad19-8e7d1fbe803a",
"run_id": "dc190557-289a-46d1-a3ce-38d6df741df2",
"snapshot_id": "d5b2be5d-0a4b-4b97-ac33-7ce02bb46b8d",
"narrative_id": "7025b0f8-283c-4741-b390-9fc7def12070",
"first_seen": "2026-10-02T17:05:18.661724+00:00",
"last_seen": "2026-10-03T05:05:23.506621+00:00",
"first_seen_at": "2026-10-02T15:51:59.916200+00:00",
"last_seen_at": "2026-10-02T16:52:11.164513+00:00",
"status": "open",
"title": "Auditing Protein-Protein Interaction Signals with Sparse Autoencoder Fingerprints",
"summary": "Protein language models have become a dominant foundation for sequence-based protein-protein interaction (PPI) prediction, but their generalization remains limited under stringent evaluation, and benchmark accuracy alone cannot reveal whether a PPI predictor learns partner-specific biological signals or exploits contextual shortcuts. Here we introduce AuditPPI, an interpretable framework that tran",
"top_tickers": [],
"why_now": [],
"metrics": {
"score_total": 0.719919,
"momentum_24h": 2,
"post_count": 2,
"unique_origin_publishers": 1,
"unique_publisher_types": 1,
"origin_share_top1": 1,
"amplifier_share_top1": null,
"duplicate_ratio": 0,
"evidence_score": 0.25,
"source_types_count": 1,
"unique_origin_domains": 1,
"unique_publishers": 1,
"quality_policy_version": 6,
"tenant_claim_coherence_policy_version": 3
},
"trend_status": "insufficient_history",
"trend_sparkline": "▁▁",
"trend_points_n": 2,
"trend_window": {
"lookback_days": 14,
"max_points": 36,
"value_key": "score_total"
},
"maturity_label": "seed",
"maturity_score": 0.24889841666666665,
"top_sources": [
{
"label": "bioRxiv: Auditing Protein-Protein Interaction Signals with Sparse Autoencoder Fingerprints",
"source_type": "rss",
"evidence_class": "primary",
"provenance_role": "origin",
"publisher_subtype": "research"
}
],
"rank_story": 0.11391235686976059,
"p_score": 0.3333333333333333,
"p_mom": 0,
"recency": 0.5889827608123758,
"risk_penalty": 0.1,
"badges": [
"seed",
"insufficient_history",
"low_evidence"
],
"storyline_category": "narrative",
"storyline_category_reason": "multi_run_continuity",
"verified_update_issuer_count": 1,
"storyline_independent_origin_count": 1,
"storyline_continuity_points": 2,
"sources": [],
"llm_title": "New preprints target protein interaction prediction and model auditing",
"llm_summary": "Two bioRxiv preprints present complementary approaches to protein–protein interaction modeling. EnsPlex integrates conformation-expanded docking with AlphaFold-Multimer, AlphaFold3 and Boltz-1, then applies quality-aware candidate selection. AuditPPI uses sparse-autoencoder-derived pair fingerprints to evaluate whether PPI models capture partner-specific signals or contextual shortcuts.",
"llm_narrative_frame": "The preprints approach protein–protein interaction prediction from complementary directions: EnsPlex expands structural candidate generation and selects models using quality-aware ranking, while AuditPPI examines whether prediction signals reflect partner-specific biology or contextual shortcuts.",
"llm_narrative_label": "Protein interaction modeling",
"llm_narrative_type": "narrative",
"llm_why_now": [
"Two new preprints address complementary bottlenecks in protein interaction prediction.",
"The studies pair structural candidate expansion with interpretability and benchmark auditing."
],
"llm_why_it_matters": [
"Broader candidate sampling could affect how predicted protein complexes are prioritized for follow-up.",
"Auditing prediction signals may help distinguish pair-specific biology from benchmark-related shortcuts.",
"Together, the studies link model generation with more rigorous evaluation of model evidence."
],
"llm_top_sources": [
{
"label": "EnsPlex preprint"
},
{
"label": "AuditPPI preprint"
}
],
"entities": {
"projects": [
"EnsPlex",
"AuditPPI"
]
},
"recurring_claims": [
{
"claim": "EnsPlex combines multiple structural generators with topology-aware candidate selection for protein-complex prediction.",
"evidence_urls": [
"https://biorxiv.org/content/10.64898/2026.09.26.754739v1?rss=1"
]
},
{
"claim": "AuditPPI uses sparse-autoencoder features to create pair fingerprints and audit the signals supporting PPI predictions.",
"evidence_urls": [
"https://biorxiv.org/content/10.64898/2026.09.27.754758v1?rss=1"
]
},
{
"claim": "The studies emphasize both broader structural candidate coverage and stricter evaluation of the evidence behind interaction predictions.",
"evidence_urls": [
"https://biorxiv.org/content/10.64898/2026.09.26.754739v1?rss=1",
"https://biorxiv.org/content/10.64898/2026.09.27.754758v1?rss=1"
]
}
],
"stance_map": [
{
"who": "EnsPlex authors",
"stance": "promotes",
"evidence_urls": [
"https://biorxiv.org/content/10.64898/2026.09.26.754739v1?rss=1"
]
},
{
"who": "AuditPPI authors",
"stance": "questions",
"evidence_urls": [
"https://biorxiv.org/content/10.64898/2026.09.27.754758v1?rss=1"
]
}
],
"quality_flags": {
"mixed_topic_risk": "low",
"promo_risk": "low",
"source_quality": "medium"
},
"editor_note": "Two fresh bioRxiv preprints address complementary challenges in protein–protein interaction modeling.",
"scope_tags": [
"protein_structure",
"protein_interactions",
"ai_for_drug_discovery",
"preprints"
],
"scope_tags_raw": [
"protein_structure",
"protein_interactions",
"ai_for_drug_discovery",
"preprints"
],
"llm_status": "accepted",
"llm_meta": {
"model": "gpt-5.6-luna",
"prompt_version": "enrich_v3",
"input_hash": "f71731329b482a722a3f739922d3532e87e696443a18dffd4cc135ee94ffe925",
"updated_at": "2026-10-03T05:08:00.534640+00:00",
"llm_metadata": {
"env": "prod",
"host": "83bc069ed062",
"stage": "signals.enrich",
"run_id": "dc190557-289a-46d1-a3ce-38d6df741df2",
"tenant": "biotech",
"service": "api",
"pipeline": "pipeline_run",
"correlation_id": "7025b0f8-283c-4741-b390-9fc7def12070"
}
},
"display_title": "New preprints target protein interaction prediction and model auditing",
"display_summary": "Two bioRxiv preprints present complementary approaches to protein–protein interaction modeling. EnsPlex integrates conformation-expanded docking with AlphaFold-Multimer, AlphaFold3 and Boltz-1, then applies quality-aware candidate selection.",
"display_tags": [
"protein_structure",
"protein_interactions",
"ai_for_drug_discovery",
"preprints"
],
"tags": [
"protein_structure",
"protein_interactions",
"ai_for_drug_discovery",
"preprints"
],
"cscope_tags": [
"protein_structure",
"protein_interactions",
"ai_for_drug_discovery",
"preprints"
],
"display_why_now": [
"Two new preprints address complementary bottlenecks in protein interaction prediction.",
"The studies pair structural candidate expansion with interpretability and benchmark auditing."
],
"display_why_it_matters": [
"Broader candidate sampling could affect how predicted protein complexes are prioritized for follow-up.",
"Auditing prediction signals may help distinguish pair-specific biology from benchmark-related shortcuts.",
"Together, the studies link model generation with more rigorous evaluation of model evidence."
],
"why_now_display": [
"Two new preprints address complementary bottlenecks in protein interaction prediction.",
"The studies pair structural candidate expansion with interpretability and benchmark auditing."
],
"why_it_matters_display": [
"Broader candidate sampling could affect how predicted protein complexes are prioritized for follow-up.",
"Auditing prediction signals may help distinguish pair-specific biology from benchmark-related shortcuts.",
"Together, the studies link model generation with more rigorous evaluation of model evidence."
],
"show_why": true,
"sources_display": [
{
"label": "biorxiv.org: AuditPPI preprint"
}
],
"editor_status": "passed",
"fields_replaced": [],
"narrative_frame_display": "The preprints approach protein–protein interaction prediction from complementary directions: EnsPlex expands structural candidate generation and selects models using quality-aware ranking, while AuditPPI examines whether prediction signals reflect partner-specific biology or contextual shortcuts.",
"display_rank_base_score": 2,
"display_rank_score": 1000000002,
"display_rank_boost": 1000000000,
"lane": "chatter",
"lane_reason": "maturity=seed",
"evidence_mix": {
"primary": 1,
"secondary": 0,
"specialist": 0,
"aggregator": 0,
"social": 0
},
"sourceStats": {
"qualityPolicyVersion": 6,
"evidenceDocumentCount": 2,
"uniqueItemCount": 2,
"uniquePublisherCount": 1,
"uniqueCanonicalOriginCount": 1,
"independentNonSocialCount": 1,
"primaryCount": 1,
"secondaryCount": 0,
"uniqueDomainsCount": 1,
"aggregatorCount": 0,
"socialCount": 0,
"sourceTypeDiversityCount": 1,
"topSourceShare": 1,
"passesTopSignalsGate": false,
"gateReason": "independentNonSocial=1; primary=1; secondary=0; documents=2; publishers=1; canonicalOrigins=1; topSourceShare=1.0; rule=>=2 unique canonical primary/secondary publishers; aggregators/social excluded"
},
"evidence_meta": {
"post_count": 2,
"unique_origin_publishers": 1,
"source_types_count": 1,
"origin_share_top1": 1
},
"gate": {
"passesTopSignalsGate": false,
"gateReason": "independentNonSocial=1; primary=1; secondary=0; documents=2; publishers=1; canonicalOrigins=1; topSourceShare=1.0; rule=>=2 unique canonical primary/secondary publishers; aggregators/social excluded"
},
"tenant_claim_coherence": {
"tenant": "biotech",
"status": "not_applicable",
"claim_scope": "not_applicable",
"reason_codes": [],
"public_anchors": {},
"evidence_anchors": {},
"debug": {
"evidence_record_count": 2,
"matching_record_count": 0,
"independent_support_count": 0,
"conflicting_record_count": 0
}
},
"evidence_quality_flag": "weak"
}
],
"next_cursor": "eyJ2IjoxLCJydW5faWQiOiJkYzE5MDU1Ny0yODlhLTQ2ZDEtYTNjZS0zOGQ2ZGY3NDFkZjIiLCJzb3J0IjoibW9tZW50dW0iLCJrIjpbMi4wLDAuNzE5OTE5LCJhZTFmNWQ1My0wYjkwLTQzMjktYWQxOS04ZTdkMWZiZTgwM2EiXX0",
"disclaimer": "No investment advice. Research signals and sources only. EarlyNarratives provides informational signals derived from public sources. It does not provide financial, legal, or tax advice."
}Capabilities
- Signals and storylines feed endpoints with filtering and rate limits
- Briefing delivery endpoints for integrations
- Evidence link payloads for auditability
Integrate in your workflow
- Route top stories into Slack or Teams for morning and evening desk updates.
- Sync storyline evidence into Notion, Airtable, or internal research trackers.
- Feed metrics into BI dashboards for momentum, concentration, and source mix monitoring.
Quick start endpoints: /v1/feed/stories, /v1/signals, /v1/storylines/search, /v1/briefings/latest.
For product access, see Pricing.