DeepXiv · CVE (cve)

About 390k public vulnerabilities (NVD, synced daily; exact count in /v1/cve corpus) plus CISA KEV exploited-in-the-wild flags and FIRST EPSS exploitation probability. Vulnerability class (closed set of 17), attacker prerequisites, affected scope, impact and one-line summaries are model-extracted. Hybrid vector + lexical search; filters and sorting in Postgres.

Corpus: 390,242 documents.

Setup

pip install -U deepxiv-sdk      # deepxiv registers a token on first use

With a registered key (https://data.rag.ac.cn/register), search and read are fair use: 60/min, 5 at once, 5,000/day; on the auto-registered token they spend the deepxiv daily limit at the credits listed below. ask draws on the daily ask allowance (resets 00:00 UTC; a site-wide daily total can run out early at peak times), or 3 lifetime trial asks on the auto-registered token. On 429, back off for Retry-After seconds. Data goes to stdout; cost, quota and sources go to stderr. Exit 0 ok, 1 API error, 2 usage error. The general CLI guide is the deepxiv-cli skill.

Start here

deepxiv fa search cve "palo alto globalprotect rce" -k 5
deepxiv fa search cve "*" -F kev=true -F severity=critical,high --sort epss -k 10   # exploited in the wild, ranked by EPSS
deepxiv fa read cve CVE-2024-3400 -l brief
deepxiv fa ask "今天有哪些高危在野利用漏洞" -d cve

Free samples: ids CVE-2024-3400; queries log4j.

Search

deepxiv fa search cve "query" [-F FILTER]... [-k N] [--offset N] [--mode M] [--sort S] [--head] [--json]

Modes: hybrid, dense, bm25, keyword. Up to 200 hits. Cost 2 (with --head: 3). Sorts: relevance, date, epss, kev_added, published (default relevance). An empty query is rejected; use "*" to list by filters only. Filters combine with AND; an undeclared field exits 2.

Advanced filters (see deepxiv fa spec cve): affected, cwe, kev_added.

Read

deepxiv fa read cve ID [ID...] [-l brief|detail|full] [-e EXTRA] [-p k=v]... [-f json|md|html]

At -l full, -p section=NAME reads one section and -p characters=N the first N characters.

Facets

deepxiv fa facets cve FIELD lists the legal values and counts, free. Use it instead of guessing a spelling. Fields: severity, vuln_class.

Ask

deepxiv fa ask "question" -d cve [-e low|medium|high] [-p k=v]... [--no-stream] [--json]

Agentic Q&A over this domain: hybrid vector + lexical search (filters the vector layer lacks fall back to Postgres full text); the model reads the hits and answers with [cve:<doc_id> <title>] citations; /stream is NDJSON with real token deltas

Always pass -d cve: automatic domain picking is still early.

What this corpus will and will not tell you

• NVD data is used under its terms: any display must carry the not-endorsed notice (see the attribution field).
• kev=true means CISA confirmed exploitation in the wild; epss is FIRST's probability estimate — different things.
• severity is derived from cvss_score; the 3,347 CVEs without a score have neither.
• Rank by exploitation risk with sort=epss, by time with sort=date.
• Two search paths: with a query, no sort and only filters the vector layer has (not affected / published / kev_added), search is hybrid vector + lexical, topped up from Postgres; otherwise, and with mode=keyword, it is Postgres full text (terms ANDed). Each hit's layer says which path.
• vuln_class / attacker_needs / impact / one_liner* are model-written — fine as a summary, not as a quote; the original text is description, in level=detail.
• Model-written fields (summarise, do not quote): affected, attacker_needs, impact, one_liner, one_liner_zh, vuln_class.
• The field table and enumerations are generated from the store's own self-description (search.field), not hand-written; enumeration counts are as of 2026-10-15.

Citing

Cite records as [CVE-2024-3400]. Say plainly when the corpus has nothing.

Generated by scripts/gen_fa_skills.py from deepxiv fa spec cve --json on 2026-10-19. Field names and prices can change; deepxiv fa spec cve is the live source.