uvx aipa-cli

aipa

uvx

1. Use plain

   uvx

   for fast cached execution:

   uvx aipa-cli deep-research

2. Fallback on failure — if a command fails with a schema or missing argument error, retry with

   @latest

   :

   uvx aipa-cli@latest deep-research

aipa

pip

pip install --upgrade aipa-cli

uvx aipa-cli

aipa

uvx

AskUserQuestion

question

aipa

AskUserQuestion

question

1. Fast Research (Recommended) — Run

   aipa deep-research

   to get the market snapshot, then you (the AI agent) replicate the multi-agent pipeline using subagents +

   aipa get-ohlcv-data

   . No API key needed. This is the recommended default — it produces thorough results without the 5-10 minute wait.
2. Full pipeline (

   --run

   ) — Run

   aipa deep-research --run

   to use the actual CLI multi-agent pipeline. Takes 5-10 minutes and requires an API key configured via

   aipa setup

   .

• If the user chooses Fast Research: follow the Fast Research: Agent-Driven Pipeline section below.
• If the user chooses Full pipeline: run

  aipa deep-research --run [QUESTION]

  and present the stdout output.

• Vietnamese stocks (

  source: vn

  ): VIC, VCB, FPT, HPG, VNM, MBB, TCB, CTG, VPB, HDB, etc.
• Cryptocurrencies (

  source: crypto

  ): BTCUSDT, ETHUSDT, BNBUSDT, SOLUSDT, etc.
• Global/Yahoo (

  source: global/yahoo

  ): AAPL, TSLA, NVDA, SPY, etc.
• SJC Gold (

  source: sjc

  ): SJC gold prices

1. Market Snapshot: Fetches latest OHLCV data for all tickers from the selected source (

   --source

   , default

   vn

   ) to identify active sectors and tickers.
2. Supervisor: Analyzes the research question and market snapshot. Decomposes into 3-5 sector-specific subtasks. Mandatory sectors depend on the source:
   • VN: Banking, Securities, Real Estate
   • Crypto: Layer 1 (BTC, ETH, SOL), DeFi, AI tokens
   • Global/Yahoo: Technology, Financials, Energy
   • SJC: Gold / Precious Metals
   Adds 0-2 additional sectors based on market activity.
3. Parallel Workers: Each worker handles one sector:
   • Fetches OHLCV data for each assigned ticker (limit=50 bars)
   • Runs volume profile for the top 3 most important tickers (30+ trading day range)
   • Fetches intraday data (1h) for tickers showing breakout/reversal patterns
   • Analyzes trend direction, VPA signals, MA score momentum, volume, and S/R
   • Cross-references volume profile levels with price action
   • Provides per-ticker assessment, sector ranking, and key risk factors
   • Workers run concurrently for efficiency
4. Aggregator: Collects all worker reports and synthesizes a unified analysis:
   • Cross-references findings across sectors
   • Builds multi-sector ranking table
   • Identifies cross-sector rotation patterns
   • Highlights key opportunities and risks
5. Reviewer: Quality assurance check:
   • Verifies no phantom stocks (tickers not in worker reports)
   • Spot-checks MA score fidelity (3-5 reported scores)
   • Confirms table completeness
   • Approves or rejects with specific feedback
   • Maximum 3 review rounds
6. Final Output: Approved report is output. If

   --output

   is specified, saved to file.

• VN: Banking, Securities, Real Estate
• Crypto: Layer 1, DeFi, AI tokens
• Global: Technology, Financials, Energy
• SJC: Gold / Precious Metals

Version gate:

aipa fundamentals

requires aipa-cli >= 0.1.43. Verify with

aipa --version

before use.

NOTE:

--lang

and

--no-system-prompt

are NOT valid for

aipa fundamentals

commands. Do NOT add them.

aipa fundamentals ratios TICKER --latest

rank

screen

aipa fundamentals rank

undefined

• Receives the sector name, ticker list, the research question, and the market snapshot
• Fetches detailed OHLCV data for its assigned tickers (

  aipa get-ohlcv-data

  with

  --limit 50

  )
• Runs

  aipa volume-profile

  for the top 3 most important tickers in its sector (highest trading value, most interesting price action, or portfolio/watchlist tickers). Use a multi-day range covering at least 30 trading days ending on today
• For tickers showing breakout/reversal patterns or key S/R levels, also fetch intraday data (

  --interval 1h --limit 50

  ) to assess entry timing
• Analyzes trend direction, VPA signals, MA momentum, volume patterns, and support/resistance
• Cross-references volume profile levels (POC, Value Area High/Low) with OHLCV analysis
• Ranks tickers within the sector
• Returns a structured sector report with per-ticker assessment, ranking, and key risk factors

• Cross-reference findings across sectors
• Build a multi-sector ranking table
• Identify cross-sector rotation patterns
• Highlight key opportunities and risks

• Verify no phantom stocks (tickers not in worker reports)
• Spot-check MA score plausibility
• Confirm table completeness
• Fix any issues and produce the final report

• aipa analyze

  : Fast, focused on 1-5 tickers, single LLM call, results in seconds.

• aipa deep-research --run

  : Covers entire sectors, multiple LLM calls across agents, results in minutes. Produces a comprehensive report with quality review.
• Fast Research (agent-driven): You replicate the pipeline with subagents. Thorough results without the CLI wait time.

deep-research

• The question spans multiple sectors
• The user wants a comprehensive report (not a quick take)
• Cross-sector comparison or rotation analysis is needed
• The user explicitly asks for "research", "deep dive", or "comprehensive analysis"

analyze

• The question is about specific tickers
• Speed matters more than depth
• The user wants a focused answer, not a full report

• stdout: The final research report. This is what you should present to the user.
• stderr: Status messages with structured markers for tracking pipeline progress.

• English: "Data by AIPriceAction | AI-powered analysis — may contain errors. Verify before trading."
• Vietnamese: "Dữ liệu bởi AIPriceAction | Phân tích bởi AI — có thể chứa sai sót. Vui lòng kiểm chứng trước khi giao dịch."

1. Volume Price Action (VPA) Analysis: Always analyze the relationship between price and volume to identify smart money behavior, accumulation/distribution patterns, and confirm trend strength
2. Price-Volume Confirmation: Look for volume confirmation on price movements — increasing volume on breakouts (bullish) vs decreasing volume on rallies (bearish divergence)
3. Wyckoff Phases: Identify market phases (Accumulation, Markup, Distribution, Markdown) based on price-volume patterns. Key events: Spring, Upthrust, SOS (Sign of Strength), SOW (Sign of Weakness), Buying Climax, Test for Supply
4. Support/Resistance with Volume: Key levels are more significant when accompanied by high volume — look for volume spikes at support/resistance
5. Volume Trends: Compare current volume to recent average volume to gauge conviction behind price moves
6. Extreme Price Changes: Detect moves exceeding ±6.7%/day (VN market limit) and search recent news/events to find causes
7. Risk Management: Every analysis must include both positive (opportunities, strengths, bullish signals) and negative (risks, weaknesses, bearish signals) insights. Quantify downside risk with specific price levels, identify what would invalidate the current thesis, and never present a one-sided view
8. Nhóm Chủ Lực (Core Market Sectors - VN Market Only): When analyzing the Vietnamese market, always contextualize tickers within their respective "Nhóm Chủ Lực" (Core Sectors) to assess systemic flow. The key groups are:
   • Nhóm Ngân hàng (Banking): VCB, BID, CTG, TCB, MBB, ACB, VPB, HDB, SHB, TPB, VIB, SSB, MSB, STB, LPB, EIB.
   • Nhóm Bất động sản (Real Estate): VIC, VHM, VRE, VPL, DIG, CEO, L14, TCH, HHS, VGC, IDC.
   • Nhóm Chứng khoán (Securities): SSI, VND, HCM, VCI, SHS, VIX, VDS.
   • Nhóm Trụ cột / Sản xuất & Bán lẻ (Blue-chips / Core Economy): HPG, HSG, NKG, FPT, MWG, GAS, GVR, PLX, BSR, MSN, VNM, SAB.
   • Nhóm Hệ sinh thái (Corporate Ecosystems):
     • Họ Vingroup: VIC, VHM, VRE, VPL.
     • Họ Bầu Thụy: STB, LPB, THD, HAG.
     • Họ Gelex ("Tuấn Mượt"): GEX, GEE, VIX, VGC, EIB, IDC.
     • Họ Hoàng Huy: TCH, HHS.
     • Họ A7: DIG, CEO, L14.
     • Họ TTC (Thành Thành Công): SBT, GEG, VDS.
     • Họ Masan: MSN, MCH, MSR, MML, VCF, VSN, NET. (Note: This classification applies only to the Vietnamese market. Crypto and Global markets do not use this specific grouping yet).

1. NEVER generate, guess, estimate, or hallucinate any numbers — prices, volumes, MA values, MA scores, percentages, dates, or any financial data. Only use data from tool results or user-provided context
2. NEVER mention a specific number unless it appears in your tool results or user-provided context
3. Use tools proactively — fetch OHLCV data before producing any analysis. The market snapshot alone is insufficient — every analysis MUST include at least one

   get-ohlcv-data

   tool call
4. When researching news or events, ALWAYS include the source name (e.g., "Source: CafeF", "Source: VNExpress")
5. Trading Hours: VN market trades 09:00–15:00 ICT (UTC+7), Mon–Fri. Crypto 24/7. If the latest bar shows unusually low volume, the session may still be in progress

This rule applies ONLY to Vietnamese stocks (

--source vn

). Crypto and global stocks are not subject to T+2 settlement.

IMPORTANT: For all stock purchases in the Vietnamese stock market, shares are only settled and available for trading (selling) on the afternoon of T+2 (specifically at 13:00 / 1:00 PM on day T+2, not T+3).

• NEVER propose or attempt to execute any Stop Loss or Take Profit actions on T+1 (the first business day after the purchase), as the shares have not yet settled.
• Always check the purchase date of any stock positions when recommending sell actions.

1. DANH_MUC.md

2. PORTFOLIO.md

3. ACCOUNT.md

1. THEO_DOI.md

2. WATCHLIST.md

1. Always check settlement status before recommending sell actions (VN stocks only) — cross-reference buy date with today. For VN stocks, shares are NOT sellable until afternoon of T+2. Does NOT apply to crypto or global stocks
2. Every analysis must quantify risk — include specific Stop Loss and Take Profit levels with reasoning, state what invalidates the thesis, calculate risk-reward ratio
3. Never present one-sided analysis — every ticker must have both bullish signals and bearish risks
4. Position sizing awareness — flag concentration risk when >30% of portfolio is in one sector
5. Daily portfolio review — mark settled positions, check SL/TP hits, flag thesis changes

• Row-by-Row Verification: When reading OHLCV data output from the CLI, you MUST strictly read the exact row for the exact date requested. Do not accidentally read data from an adjacent row or a different ticker's block in multi-ticker outputs.
• Precision Filter with Grep: To minimize reading errors and context volume, always use

  grep -E

  to isolate your target dates across one or multiple tickers. Use

  "time"

  as your header anchor.
  • Surgical view (Header + Today + Breakout Day):

    uvx aipa-cli get-ohlcv-data TCB MSB STB | grep -E "time|2026-05-27|2026-05-07"

  • Comparing recent days:

    uvx aipa-cli get-ohlcv-data VND | grep -E "time|2026-05-27|2026-05-26"

• Explicit Value Comparison: Before concluding whether a trend is broken or intact, explicitly state the values being compared:

  [Close Price]

  vs

  [MA/EMA Value]

  .
  • Example: "Close is 17.750, EMA20 is 16.881. 17.750 > 16.881 → Price is ABOVE EMA20 (Trend intact)."
• Breakout Validation: A breakout (significant positive price change + high volume) creates a critical support at the structural breakout level — the top of the pre-breakout base/range, the prior swing high, or the pattern's neckline. The breakout candle's Low is NOT a reliable invalidation point: it can extend well below the structural level due to gap opens, intraday noise, or volatile entry bars.
  • The correct invalidation is a fall back below the structural breakout level, not below the candle's Low.
  • If price pulls back but stays above the structural level, the breakout is intact — this is a healthy pullback.
  • If price falls below the structural breakout level, it is a Failed Breakout / Structural Violation.
  • Action: Always identify the pre-breakout structure first. Only then assess whether a pullback is healthy (above structure) or a failure (below structure).

1. Fast Research is the default: Unless the user explicitly asks for

   --run

   , run

   aipa deep-research

   to get the snapshot, then orchestrate the pipeline yourself (Step 1-5 above).
2. Be specific with the question: The default question produces a general market overview. For better results, provide a specific research question targeting sectors or themes of interest.
3. Save reports: Use

   --output FILE

   with

   --run

   to save the report. This is useful for sharing or later reference.
4. Resume on failure: If a

   --run

   session is interrupted, the session ID is printed in the output. Use

   --resume ID

   to continue.
5. Vietnamese analysis: Add

   --lang vn

   when the user communicates in Vietnamese or wants Vietnamese output. For Fast Research, write the report in the user's language.
6. Combine with analyze: For a complete workflow, use deep research for the broad picture, then

   analyze

   for deep dives into specific tickers identified in the research.
7. The pipeline self-corrects: The reviewer may reject the aggregator's output and request revisions. This is normal and ensures quality. The pipeline retries up to 3 times.
8. Mandatory sectors depend on source: Each source has its own set of mandatory sectors (e.g., VN uses Banking/Securities/Real Estate, crypto uses Layer 1/DeFi/AI tokens, global uses Technology/Financials/Energy). The supervisor adds 0-2 more sectors based on market conditions.
9. Auto-uppercase: Ticker symbols in questions are automatically processed. The pipeline handles uppercase conversion internally.
10. Use

    --source

    for non-VN markets: Add

    --source crypto

    ,

    --source global

    , or

    --source sjc

    to research other markets. The supervisor, workers, tools, and default question all adapt to the selected source automatically.
11. Use

    aipa performers

    to identify leaders and laggards — run multiple perspectives: Before or during research, run

    aipa performers

    with multiple

    --sort-by

    values. Always run at least these two: default (price change) and value (trading value). Add MA scores for trend context. Cross-referencing the lists helps you pick better tickers for sector workers and provides richer ranking context for the final report.
    bash

    aipa performers                                          # price change — top gainers / worst losers
    aipa performers --sort-by value                          # trading value — where the money flows
    aipa performers --sort-by ma50_score                     # MA50 trend — strongest/weakest medium-term trends
    aipa performers --sort-by ma20_score                     # MA20 trend — strongest/weakest short-term trends
    aipa performers --sort-by total_money_changed            # money flow change — unusual capital activity
    aipa performers --source crypto --sort-by value          # crypto by trading value
    aipa performers --group NGAN_HANG --sort-by value        # banking sector by trading value
    aipa performers --group CHUNG_KHOAN --sort-by close_changed  # securities sector top gainers

12. Use

    aipa volume-profile

    for key price levels: When workers need support/resistance context for a sector, run

    aipa volume-profile TICKER

    on the top movers to get POC, value area, and volume-weighted statistics. Prefer multi-day ranges over single-day profiles — they produce more reliable support/resistance levels and smooth out intraday noise. Use

    --start-date

    and

    --end-date

    covering at least 20 trading days as the default approach. Only use a single

    --date

    when the user explicitly asks for one specific day. Examples:

    • aipa volume-profile VCB --start-date 2026-04-14 --end-date 2026-05-09

      — 1-month range (preferred default)

    • aipa volume-profile VCB --start-date 2026-04-28 --end-date 2026-05-09 --bins 30

      — 2-week range

    • aipa volume-profile BTCUSDT --source crypto --bins 30 --start-date 2026-05-05 --end-date 2026-05-09

      — crypto multi-day