Skip to main content

Astraea ADK (LegionADK)

Enterprise-Grade Mixture-of-Agents (MoA) Orchestration & Autonomous Agent Development Kit

Python Version Pydantic Architecture Status API Documentation License

Astraea ADK (LegionADK) adalah framework dan runtime tingkat enterprise untuk perancangan, orkestrasi, tata kelola keamanan (governance), memori graf lokal, serta komunikasi sistem multi-agen berbasis Mixture-of-Agents (MoA). ADK ini dirancang dari nol dengan fokus pada determinisme matematis, pencegahan balapan data (race condition), efisiensi biaya inferensi token LLM, dan keandalan tingkat produksi (production-grade resilience).


📑 Daftar Isi


🏗 Arsitektur Sistem (Layer Breakdown)

Astraea ADK System Architecture

Astraea ADK dibangun dengan arsitektur berlapis (layered architecture) yang modular, deterministik, dan terisolasi dengan batas tanggung jawab (separation of concerns) yang jelas. Setiap lapisan saling berkolaborasi melalui protokol komunikasi yang terstandarisasi:

1. Protocol Layer (MoA Orchestration Engine)

Lapisan orkestrasi tingkat tinggi yang mengendalikan interaksi, kolaborasi, dan alur kerja antar-agen:

  • MoA Orchestration: Mengelola eksekusi Mixture-of-Agents secara adaptif. Agen koordinator dapat menganalisis persoalan secara mandiri, menentukan pembagian peran, dan membagi tugas ke berbagai spesialis terkait.
  • Multi-Agent Debate: Memfasilitasi pertukaran ide, sintesis kompromi, dan evaluasi silang (cross-critique) multi-putaran antar-spesialis sebelum draf akhir disatukan kembali oleh peran agregator.
  • Strict Chain-of-Command: Mengatur hierarki delegasi berjenjang (atasan ke bawahan) dengan batas kedalaman eksekusi terkontrol, pelacakan tumpukan panggilan (call stack tracking), serta pencegahan siklus delegasi (anti-cycle protection).

2. Runtime Layer (10-Phase Agent Turn Engine)

Mesin eksekusi inti yang memandu siklus hidup giliran (turn lifecycle) setiap agen secara deterministik dan terukur:

  • 10-Phase Execution Pipeline: Menjalankan setiap giliran interaksi melalui tahapan terstandarisasi—mulai dari audit tata kelola awal (preflight), perakitan prompt dinamis, pemanggilan model terawasi (watchdog invocation), eksekusi aksi, evaluasi keamanan pasca-eksekusi, hingga finalisasi status.
  • Context Compaction & Token Budgeting: Pengelolaan kuota token proaktif dengan strategi reduksi riwayat dua tingkat (summarization & eviction), menjaga relevansi konteks percakapan panjang tanpa melampaui jendela konteks model.
  • Immutable State Transitions: Menjamin seluruh manipulasi status percakapan bersifat beku (immutable frozen state), mencegah mutasi data yang tidak terduga antar-giliran.

3. Gateway Layer (Resilient I/O & Routing)

Gerbang komunikasi berkecepatan tinggi yang menjembatani runtime agen dengan berbagai penyedia model LLM (foundation models):

  • Model Router & Load Balancer: Manajemen lalu lintas inferensi multi-penyedia dengan isolasi kuota RPM/TPM, failover cascade otomatis, dan pemulihan mandiri (auto-cooldown) saat terjadi batasan laju (rate-limiting).
  • Byte-Stable Cache Boundary: Segmentasi prompt statis dan dinamis dengan verifikasi hash deterministik untuk memastikan efisiensi biaya tertinggi dan retensi prompt caching penyedia LLM (>90% Cache HIT).
  • Multi-LLM Adapters: Abstraksi konektor universal untuk penyedia model cloud terkemuka maupun model lokal (self-hosted / private inference).
  • Self-Healing Tool Repair: Mekanisme penanganan anomali otomatis yang mampu memperbaiki payload JSON yang terpotong atau mengalami inkonsistensi sintaks sebelum diproses lebih lanjut.

4. Governance Layer (Security & Safety Guardrails)

Sistem pertahanan berlapis (defense-in-depth) yang menjamin keamanan dan kepatuhan eksekusi di lingkungan produksi:

  • Deterministic Verifier: Validasi parametrik ketat terhadap seluruh aksi agen sebelum dieksekusi, mencegah manipulasi batas direktori kerja (path traversal) dan akses resource ilegal.
  • Smart Guardian: Mesin audit perilaku agen secara real-time untuk mengevaluasi risiko eksekusi kode berbahaya, manipulasi instruksi (prompt injection), dan pola loop anomali.
  • Circuit Breakers: Sakelar proteksi otomatis yang memutus rantai eksekusi saat layanan hilir mengalami kegagalan berulang, mencegah pemborosan kuota dan kelelahan sistem.
  • Parallel Execution Guards: Pembatasan hak akses operasi mutasi bagi agen yang berjalan serentak, menjamin integritas data dan eliminasi total terhadap kondisi balapan (race condition).

5. Memory Layer (Local Knowledge Graph & Vault)

Infrastruktur memori graf pengetahuan lokal tanpa ketergantungan pada database eksternal yang rumit:

  • Knowledge Graph Vault: Penyimpanan dokumen pengetahuan berbasis format terbuka dengan metadata terstruktur dan transaksi berkas atomik yang aman (atomic file-locking).
  • Wiki-Links & Semantic Relations: Pengindeksan otomatis relasi dua arah (bidirectional links) dan penelusuran relasi dokumen secara deterministik untuk menemukan konteks yang relevan secara instan.

🌟 Fitur Utama (Core Features)

Astraea ADK dirancang untuk memenuhi standar keandalan, keamanan, dan efisiensi tingkat produksi (enterprise-grade) bagi ekosistem multi-agen modern:

  • 100% Role-Agnostic Multi-Agent Ecosystem Tidak ada peran atau persona yang di-hardcode. Pengembang memiliki kebebasan penuh merancang struktur tim agen, persona (soul), SOP operasional, serta wewenang alat secara dinamis sesuai kebutuhan spesifik domain industri.

  • Advanced Mixture-of-Agents (MoA) Collaboration Mendukung kolaborasi multi-agen canggih melalui dua pola operasional: konsensus debat multi-putaran untuk eksplorasi dan sintesis ide, serta triase otonom di mana koordinator memecah dan mendelegasikan tugas ke sub-tim spesialis secara mandiri.

  • Strict Chain-of-Command & Anti-Cycle Hierarchies Mendukung pendelegasian hierarkis berjenjang (atasan ke bawahan) dengan batas kedalaman terkontrol, jaminan zero double-call pada LLM untuk efisiensi token, serta perlindungan tumpukan panggilan (call stack guard) yang mencegah siklus delegasi tanpa akhir.

  • Byte-Stable Prompt Caching & Cost Optimizer Optimalisasi biaya inferensi tingkat lanjut dengan memisahkan identitas agen statis dari konteks dinamis secara deterministik. Memaksimalkan rasio Cache HIT hingga >90% pada penyedia LLM modern, memangkas latensi dan biaya operasional secara signifikan.

  • Multi-Provider Resilience & Auto-Failover Konektivitas terintegrasi ke berbagai penyedia model cloud terkemuka serta infrastruktur self-hosted / private proxy. Dilengkapi isolasi kuota RPM/TPM mandiri, mekanisme failover cascade, serta pemulihan otonom saat terkena batas laju (HTTP 429/5xx).

  • Self-Healing Tool Calling & JSON Repair Mesin perbaikan otomatis yang secara mandiri menambal keluaran model yang terpotong oleh batasan token (max_tokens), menyeimbangkan tanda kurung, membersihkan format data, dan menjamin keberhasilan pemanggilan fungsi tanpa crash.

  • Enterprise-Grade Governance & Multi-Layer Guardrails Validasi deterministik sebelum aksi (pre-execution verifier), audit keamanan perilaku dan analisis risiko kode secara real-time, sakelar pemutus sirkuit (circuit breaker), serta isolasi ketat terhadap sistem operasi dan direktori berkas.

  • Concurrency & Race-Condition Immunity Eksekusi multi-agen paralel yang kebal terhadap konflik data berkat mekanisme isolasi hak akses mutasi pada agen paralel serta transaksi berkas atomik (deadlock-free file locking).

  • Local Knowledge Graph Memory Engine Penyimpanan memori jangka panjang berbasis graf pengetahuan lokal dengan penautan semantik dua arah (Wiki-Links) dan penelusuran jalur keterkaitan dokumen yang cepat, sepenuhnya mandiri tanpa dependensi database eksternal yang kompleks.


⚙️ Konfigurasi Lingkungan (.env)

Buat berkas .env pada direktori root proyek untuk menghubungkan penyedia LLM Anda:

# ============================================================================
# Universal Custom Provider (Aggregator / Private Proxy / Self-hosted vLLM)
# ============================================================================
CUSTOM_PROVIDER_API_KEY=your_api_key_here
CUSTOM_PROVIDER_BASE_URL=https://api.your-provider.com/v1

# Parameter Model Utama
TEST_PRIMARY_PROVIDER=custom
TEST_PRIMARY_MODEL=deepseek-v4.1-flash

# ============================================================================
# Official Providers (Opsional - Sesuai Kebutuhan)
# ============================================================================
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GEMINI_API_KEY=AIzaSy...
OLLAMA_BASE_URL=http://localhost:11434/v1
VLLM_BASE_URL=http://localhost:8000/v1

# Matikan mode live pada test suite jika hanya ingin menguji algoritma offline:
# LEGION_SKIP_LIVE=1

🚀 Panduan Penggunaan Cepat (Quickstart)

A. Membangun Roster Agen Dinamis

LegionADK bersifat role-agnostic. Anda bebas menentukan nama peran, persona soul, dan batasan SOP:

from LegionADK.core.llm_spec import LLMSpec
from LegionADK.protocol import AgentRoster

roster = AgentRoster([
    LLMSpec(
        handle="@coordinator",
        agent_name="Koordinator Utama",
        soul_prompt="Kamu adalah koordinator tim yang analitis dan cermat.",
        system_prompt="Analisis topik dan delegasikan tugas ke spesialis terkait.",
        tool_permissions=["comm:delegate"],
    ),
    LLMSpec(
        handle="@architect",
        agent_name="Software Architect",
        soul_prompt="Kamu mengutamakan modularitas, skalabilitas, dan keandalan sistem.",
        system_prompt="Berikan rekomendasi arsitektur sistem yang konkret.",
        tool_permissions=["comm:delegate"],
    ),
    LLMSpec(
        handle="@security",
        agent_name="Security Specialist",
        soul_prompt="Kamu fokus pada keamanan memori, enkripsi, dan audit risiko.",
        system_prompt="Identifikasi kerentanan dan rekomendasikan kontrol pengamanan.",
    ),
])

B. Cara 1: Diskusi Multi-Putaran Eksplisit (run_moa_debate)

Gunakan saat pengguna meminta beberapa agen berdiskusi mencari titik kompromi:

import asyncio
from LegionADK.core.llm_spec import contract_message
from LegionADK.protocol import OrchestrationPolicy, run_moa_debate

async def main():
    env = contract_message(
        sender="@user",
        target="@coordinator",
        content="Silakan @architect berdiskusi dengan @security untuk merancang sistem autentikasi perbankan.",
    )

    policy = OrchestrationPolicy(max_discussion_rounds=2)
    result = await run_moa_debate(
        envelope=env,
        roster=roster,
        aggregator_handle="@coordinator",
        policy=policy,
    )

    print(f"Status Orkestrasi: {result.status}")
    print(f"Sintesis Akhir:\n{result.final_content}")

asyncio.run(main())

C. Cara 2: Triase Otonom & Pendelegasian Mandiri (run_autonomous_triage)

Koordinator mengevaluasi topik. Jika sederhana, koordinator langsung menjawab (jalur cepat STANDALONE). Jika membutuhkan masukan tim, koordinator memanggil spesialis secara mandiri lalu menyintesiskan hasilnya:

from LegionADK.protocol import run_autonomous_triage

async def main():
    env = contract_message(
        sender="@user",
        target="@coordinator",
        content="Kita mengalami lonjakan trafik transaksi 500% malam ini. Analisis dan buat rencana mitigasi.",
    )

    result = await run_autonomous_triage(
        envelope=env,
        roster=roster,
        coordinator_handle="@coordinator",
    )

    print(f"Mode Terpilih : {result.mode}") # AUTONOMOUS_TRIAGE atau STANDALONE
    print(f"Agen Spesialis: {result.proposer_handles}")
    print(f"Keputusan Akhir:\n{result.final_content}")

asyncio.run(main())

D. Fasad Tingkat Atas: Router Otomatis (dispatch_envelope)

Pintu masuk terpadu yang otomatis menganalisis amplop dan memilih alur eksekusi terbaik:

from LegionADK.protocol import dispatch_envelope

async def handle_incoming_message(envelope):
    # Otomatis mendeteksi:
    # 1. AUTONOMOUS_TRIAGE (jika target adalah koordinator)
    # 2. EXPLICIT_MOA (jika target >= 2 agen atau menyebut banyak @handle)
    # 3. STANDALONE (jika target 1 agen biasa)
    return await dispatch_envelope(envelope=envelope, roster=roster)

E. Strict Chain-of-Command Rekursif (Hierarki Berjenjang)

Mendukung pendelegasian hierarkis berlapis: User $\to$ @coordinator $\to$ @architect $\to$ @security $\to$ @architect $\to$ @coordinator $\to$ User.

policy = OrchestrationPolicy(
    max_delegation_depth=3,      # Batas kedalaman hierarki (mencegah runaway recursion)
    max_discussion_rounds=1,     # Putaran debat per level
)

result = await dispatch_envelope(
    envelope=env,
    roster=roster,
    policy=policy,
    triage_handles=["@coordinator"],
)

# Seluruh prompt, cached, dan completion tokens dari @security terakumulasi ke @architect
# dan teragregasi secara presisi ke dalam total @coordinator.
print(f"Total Token Tim: {result.total_prompt_tokens} prompt, {result.total_completion_tokens} completion")

🧪 Verifikasi & Checkpoint Test Suite

Seluruh kapabilitas LegionADK diverifikasi melalui rangkaian pengujian komprehensif dari Checkpoint 1 hingga Checkpoint 4.

Menjalankan Seluruh Pengujian Checkpoint 4

# Jalankan seluruh pengujian (O1 - O12) termasuk live LLM test:
$env:PYTHONPATH="."
python checkpoint_4_orchestrator.py

# Atau jalankan mode offline murni (melewati live LLM):
$env:LEGION_SKIP_LIVE="1"
python checkpoint_4_orchestrator.py

Hasil Ringkasan Checkpoint 4 (100% PASS)

Kode Kategori Nama Pengujian Status Rangkuman Hasil
O1 Offline Parser @handle & Code-Fence PASS Email (kontak@pm.com) & dekorator kode diabaikan; urutan kemunculan terjaga deterministik.
O2 Offline Resolusi Roster & @all PASS Ekspansi @all otomatis mengecualikan pengirim untuk mencegah self-mention ping-pong.
O3 Offline Parallel Mutation Guard PASS Proposer paralel diblokir saat mencoba mengeksekusi tool mutasi Tier 1+ (workspace:write_config).
O4 Offline Isolasi Kegagalan Proposer PASS 1 Proposer timeout/gagal tidak menggagalkan sintesis Aggregator (graceful degradation).
O5 Offline Cara 1 (Debat Multi-Putaran) PASS 2 putaran diskusi Proposer terkirim via dynamic_context byte-stable dan disintesis sempurna.
O6 Offline Cara 2 (Triase Otonom Koordinator) PASS Koordinator secara mandiri mengevaluasi direktori peran, memilih spesialis, dan menyintesis solusi.
O9 Offline Recursive Strict Chain-of-Command PASS Rantai 3-tingkat terdelegasi mulus dengan pembuktian Zero Double-Triage dan agregasi token akurat.
O10 Offline Governance Cutoff & Anti-Cycle PASS max_delegation_depth=1 menahan rekursi berlebih; deteksi siklus call stack menghasilkan [SIKLUS TERDETEKSI].
O7 Live LLM Live Cara 1 (Debat Eksplisit) PASS Eksekusi nyata API: @cto dan @pm berdiskusi mencari titik kompromi, disintesis oleh @supervisor.
O8 Live LLM Live Cara 2 (Triase Otonom) PASS Eksekusi nyata API: @cos mandiri menganalisis lonjakan biaya 300% dan memanggil tim teknis.
O11 Live LLM Live Recursive Chain-of-Command PASS Eksekusi nyata API: Rantai hierarkis User $\to$ @cos $\to$ @cto $\to$ @pm tuntas disintesis berjenjang.
O12 Live LLM Live Governance Cutoff PASS Eksekusi nyata API: max_delegation_depth=1 berhasil ditegakkan tanpa memicu sub-triase liar.

🔒 Lisensi

Proyek ini dilindungi di bawah lisensi Proprietary and Confidential License. Hak Cipta © 2025 Astraea ADK Contributors. Seluruh hak cipta dilindungi undang-undang.

Dilarang keras menyebarluaskan, mendistribusikan ulang, atau mereproduksi kode sumber ini tanpa izin tertulis eksplisit dari pemilik hak cipta.

Metadata

Release files for astraea-adk 1.5.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Built distributions (wheels)

Table of built distributions (wheels) for astraea-adk 1.5.0
File Actions
astraea_adk-1.5.0-cp315-cp315-win_amd64.whl CPython 3.15 CPython 3.15 Windows x86-64 Details for astraea_adk-1.5.0-cp315-cp315-win_amd64.whl
astraea_adk-1.5.0-cp315-cp315-manylinux2014_x86_64.whl CPython 3.15 CPython 3.15 Linux glibc 2.17+ x86-64 Details for astraea_adk-1.5.0-cp315-cp315-manylinux2014_x86_64.whl
astraea_adk-1.5.0-cp315-cp315-macosx_10_15_universal2.whl CPython 3.15 CPython 3.15 macOS 10.15+ universal2 (ARM64, x86-64) Details for astraea_adk-1.5.0-cp315-cp315-macosx_10_15_universal2.whl
astraea_adk-1.5.0-cp314-cp314-win_amd64.whl CPython 3.14 CPython 3.14 Windows x86-64 Details for astraea_adk-1.5.0-cp314-cp314-win_amd64.whl
astraea_adk-1.5.0-cp314-cp314-manylinux2014_x86_64.whl CPython 3.14 CPython 3.14 Linux glibc 2.17+ x86-64 Details for astraea_adk-1.5.0-cp314-cp314-manylinux2014_x86_64.whl
astraea_adk-1.5.0-cp314-cp314-macosx_10_15_universal2.whl CPython 3.14 CPython 3.14 macOS 10.15+ universal2 (ARM64, x86-64) Details for astraea_adk-1.5.0-cp314-cp314-macosx_10_15_universal2.whl
astraea_adk-1.5.0-cp313-cp313-win_amd64.whl CPython 3.13 CPython 3.13 Windows x86-64 Details for astraea_adk-1.5.0-cp313-cp313-win_amd64.whl
astraea_adk-1.5.0-cp313-cp313-manylinux2014_x86_64.whl CPython 3.13 CPython 3.13 Linux glibc 2.17+ x86-64 Details for astraea_adk-1.5.0-cp313-cp313-manylinux2014_x86_64.whl
astraea_adk-1.5.0-cp313-cp313-macosx_10_13_universal2.whl CPython 3.13 CPython 3.13 macOS 10.13+ universal2 (ARM64, x86-64) Details for astraea_adk-1.5.0-cp313-cp313-macosx_10_13_universal2.whl
astraea_adk-1.5.0-cp312-cp312-win_amd64.whl CPython 3.12 CPython 3.12 Windows x86-64 Details for astraea_adk-1.5.0-cp312-cp312-win_amd64.whl
astraea_adk-1.5.0-cp312-cp312-manylinux2014_x86_64.whl CPython 3.12 CPython 3.12 Linux glibc 2.17+ x86-64 Details for astraea_adk-1.5.0-cp312-cp312-manylinux2014_x86_64.whl
astraea_adk-1.5.0-cp312-cp312-macosx_10_13_universal2.whl CPython 3.12 CPython 3.12 macOS 10.13+ universal2 (ARM64, x86-64) Details for astraea_adk-1.5.0-cp312-cp312-macosx_10_13_universal2.whl
astraea_adk-1.5.0-cp311-cp311-win_amd64.whl CPython 3.11 CPython 3.11 Windows x86-64 Details for astraea_adk-1.5.0-cp311-cp311-win_amd64.whl
astraea_adk-1.5.0-cp311-cp311-manylinux2014_x86_64.whl CPython 3.11 CPython 3.11 Linux glibc 2.17+ x86-64 Details for astraea_adk-1.5.0-cp311-cp311-manylinux2014_x86_64.whl
astraea_adk-1.5.0-cp311-cp311-macosx_10_9_universal2.whl CPython 3.11 CPython 3.11 macOS 10.9+ universal2 (ARM64, x86-64) Details for astraea_adk-1.5.0-cp311-cp311-macosx_10_9_universal2.whl
astraea_adk-1.5.0-cp310-cp310-win_amd64.whl CPython 3.10 CPython 3.10 Windows x86-64 Details for astraea_adk-1.5.0-cp310-cp310-win_amd64.whl
astraea_adk-1.5.0-cp310-cp310-manylinux2014_x86_64.whl CPython 3.10 CPython 3.10 Linux glibc 2.17+ x86-64 Details for astraea_adk-1.5.0-cp310-cp310-manylinux2014_x86_64.whl
astraea_adk-1.5.0-cp310-cp310-macosx_10_9_universal2.whl CPython 3.10 CPython 3.10 macOS 10.9+ universal2 (ARM64, x86-64) Details for astraea_adk-1.5.0-cp310-cp310-macosx_10_9_universal2.whl

Total release size: 80.6 MB

Release files / astraea_adk-1.5.0-cp315-cp315-win_amd64.whl

Download URL astraea_adk-1.5.0-cp315-cp315-win_amd64.whl
Size 1.4 MB
Tags CPython 3.15 Windows x86-64
SHA-256 checksum
How to use checksums
6ec2e0cc3478be7089cbc61ebee8c62862e392eaa8a16393d310b262299ae5ed
BLAKE2b-256 checksum
How to use checksums
f6f52d98c5abba52a1fef5ea2039a11726e5800612456ebcd3727b45ae535a24
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

Release files / astraea_adk-1.5.0-cp315-cp315-manylinux2014_x86_64.whl

Download URL astraea_adk-1.5.0-cp315-cp315-manylinux2014_x86_64.whl
Size 9.1 MB
Tags CPython 3.15 Linux glibc 2.17+ x86-64
SHA-256 checksum
How to use checksums
006ad868f8d8de769814eac183032680ba96b9f2f8094e818c289758dd32d4c0
BLAKE2b-256 checksum
How to use checksums
4820ad6920b0400ddcd84d3b47f715ee0d9c8041b8900e67423f8574267f582d
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

Release files / astraea_adk-1.5.0-cp315-cp315-macosx_10_15_universal2.whl

Download URL astraea_adk-1.5.0-cp315-cp315-macosx_10_15_universal2.whl
Size 2.9 MB
Tags CPython 3.15 macOS 10.15+ universal2 (ARM64, x86-64)
SHA-256 checksum
How to use checksums
10f19583cd1f22ef90c7364c095d4794f49b0bf239700891720f46e29dd33a3c
BLAKE2b-256 checksum
How to use checksums
c9e48df73c9aafdc034531c8e634859f7d76e4eedb2688442744c41bd42a69ff
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

Release files / astraea_adk-1.5.0-cp314-cp314-win_amd64.whl

Download URL astraea_adk-1.5.0-cp314-cp314-win_amd64.whl
Size 1.4 MB
Tags CPython 3.14 Windows x86-64
SHA-256 checksum
How to use checksums
b94ddade6f55a328d5c206e4547a30517e80a70646ee111ba1fedd64ebb3218b
BLAKE2b-256 checksum
How to use checksums
da197c6ffaed0f21ea13ab45fe7d16a672f1cc7492abb7fb5c5babe62f51cd85
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

Release files / astraea_adk-1.5.0-cp314-cp314-manylinux2014_x86_64.whl

Download URL astraea_adk-1.5.0-cp314-cp314-manylinux2014_x86_64.whl
Size 9.1 MB
Tags CPython 3.14 Linux glibc 2.17+ x86-64
SHA-256 checksum
How to use checksums
737350205757740d40f40c2a4d00bcbbc10c93e331ef32c09945f64ddd1fc4c8
BLAKE2b-256 checksum
How to use checksums
15d2c5ffd1ffe3b61061d38a9a574e1d1aa85481ae7ca53b95f871a5e005469d
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

Release files / astraea_adk-1.5.0-cp314-cp314-macosx_10_15_universal2.whl

Download URL astraea_adk-1.5.0-cp314-cp314-macosx_10_15_universal2.whl
Size 2.9 MB
Tags CPython 3.14 macOS 10.15+ universal2 (ARM64, x86-64)
SHA-256 checksum
How to use checksums
a44631c5d7b6a8f7f06118758df4a15e48860b99f696a08c54a440af5b26620f
BLAKE2b-256 checksum
How to use checksums
7efaf13bff68f12381bc326e48ce0a54333d6f98a82bb2767b70be32e5207274
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

Release files / astraea_adk-1.5.0-cp313-cp313-win_amd64.whl

Download URL astraea_adk-1.5.0-cp313-cp313-win_amd64.whl
Size 1.3 MB
Tags CPython 3.13 Windows x86-64
SHA-256 checksum
How to use checksums
bdd31485fd674cc2ece21cfa29f9b5715e9e71f0c5b1923bbc3f7d340861ff3c
BLAKE2b-256 checksum
How to use checksums
007d717d2237edfecd4cf438d81db5fa71ed6f4f30771811ac7d89b9252b2d29
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

Release files / astraea_adk-1.5.0-cp313-cp313-manylinux2014_x86_64.whl

Download URL astraea_adk-1.5.0-cp313-cp313-manylinux2014_x86_64.whl
Size 9.3 MB
Tags CPython 3.13 Linux glibc 2.17+ x86-64
SHA-256 checksum
How to use checksums
6f58e4f9142f56083e2096e266f97e1047d290e3e6fa6ec6a6c2b3accffe2328
BLAKE2b-256 checksum
How to use checksums
4d81aa7b2e35ac7c7c0d59f6adc8e6fba3d8f178f9d5dc8c171109bf7d436e07
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

Release files / astraea_adk-1.5.0-cp313-cp313-macosx_10_13_universal2.whl

Download URL astraea_adk-1.5.0-cp313-cp313-macosx_10_13_universal2.whl
Size 2.9 MB
Tags CPython 3.13 macOS 10.13+ universal2 (ARM64, x86-64)
SHA-256 checksum
How to use checksums
64b7956a0b97d154fc57a618d6c3a65d6618cfb5ffdef9eeb3ec667750738027
BLAKE2b-256 checksum
How to use checksums
304d93246ce87d46a660f297b30860c7670def45da12b761399985cd0c51cff0
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

Release files / astraea_adk-1.5.0-cp312-cp312-win_amd64.whl

Download URL astraea_adk-1.5.0-cp312-cp312-win_amd64.whl
Size 1.3 MB
Tags CPython 3.12 Windows x86-64
SHA-256 checksum
How to use checksums
4b143987932711db65f735ef4b7564298521a00f427aabcb1b4a19d75b72ae17
BLAKE2b-256 checksum
How to use checksums
3ef3e96eb37c84c16305d1c4327b50f4b492e6819c6a5c9c804a9e9d6564cc97
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

Release files / astraea_adk-1.5.0-cp312-cp312-manylinux2014_x86_64.whl

Download URL astraea_adk-1.5.0-cp312-cp312-manylinux2014_x86_64.whl
Size 9.4 MB
Tags CPython 3.12 Linux glibc 2.17+ x86-64
SHA-256 checksum
How to use checksums
ca010164a8ca3a8a23d457b63680c7d5e94bdc23136f74a29d513fa6d414600d
BLAKE2b-256 checksum
How to use checksums
80c19a5845bbb75ad8e3847a90c980e2283bfa7a34f9b23d5295952c83133d1d
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

Release files / astraea_adk-1.5.0-cp312-cp312-macosx_10_13_universal2.whl

Download URL astraea_adk-1.5.0-cp312-cp312-macosx_10_13_universal2.whl
Size 2.9 MB
Tags CPython 3.12 macOS 10.13+ universal2 (ARM64, x86-64)
SHA-256 checksum
How to use checksums
3a7cc651398139c3869ea37860c6b3a002a427ea23f47ec8e84d00808d25bc4d
BLAKE2b-256 checksum
How to use checksums
9895fd3ed883d188d2313667ff4c97e64d1910ef3e8ecffe32a4ea2a1a8b86c1
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

Release files / astraea_adk-1.5.0-cp311-cp311-win_amd64.whl

Download URL astraea_adk-1.5.0-cp311-cp311-win_amd64.whl
Size 1.4 MB
Tags CPython 3.11 Windows x86-64
SHA-256 checksum
How to use checksums
6182c1e4f7347141086a98d2e9e750b20339eaaa1ea8fc12f5f14cd8994280ce
BLAKE2b-256 checksum
How to use checksums
723ba5df790176fa0bcbf9afa0ac21fc25e9497744a22d737fd515099345bbac
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

Release files / astraea_adk-1.5.0-cp311-cp311-manylinux2014_x86_64.whl

Download URL astraea_adk-1.5.0-cp311-cp311-manylinux2014_x86_64.whl
Size 9.2 MB
Tags CPython 3.11 Linux glibc 2.17+ x86-64
SHA-256 checksum
How to use checksums
d0e77c36acb078ff11f769992886dfddb07b4b2698dedadb66b3d9ef60d29beb
BLAKE2b-256 checksum
How to use checksums
10d2ae5b075eb58cad1e4bc9effd2a178c2656bdd2b6c1ee087b1bca6ebb6bb0
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

Release files / astraea_adk-1.5.0-cp311-cp311-macosx_10_9_universal2.whl

Download URL astraea_adk-1.5.0-cp311-cp311-macosx_10_9_universal2.whl
Size 2.9 MB
Tags CPython 3.11 macOS 10.9+ universal2 (ARM64, x86-64)
SHA-256 checksum
How to use checksums
b6e7352c0843ae16cc58ba3639254dcba808b20dbdf3ed076b32d709da7d5d7e
BLAKE2b-256 checksum
How to use checksums
93e920721372571fe577ada7ab98f78c54fff346a11eacba03c7cd60e5ccfd32
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

Release files / astraea_adk-1.5.0-cp310-cp310-win_amd64.whl

Download URL astraea_adk-1.5.0-cp310-cp310-win_amd64.whl
Size 1.4 MB
Tags CPython 3.10 Windows x86-64
SHA-256 checksum
How to use checksums
2a0b0bdb583f2fa5bf7cca98c36bccb9e23da33c9d937072503c215666f5e7ee
BLAKE2b-256 checksum
How to use checksums
4e8b923d905f8f91aca4d0d3dcfdbe1216754932cdaf44a24ef6b98661cf845e
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

Release files / astraea_adk-1.5.0-cp310-cp310-manylinux2014_x86_64.whl

Download URL astraea_adk-1.5.0-cp310-cp310-manylinux2014_x86_64.whl
Size 8.9 MB
Tags CPython 3.10 Linux glibc 2.17+ x86-64
SHA-256 checksum
How to use checksums
ece3c8bfee6d65ce421f699ab1953f380a8527fa2a1520d4b0f6ef1771be5354
BLAKE2b-256 checksum
How to use checksums
fccde717032f4d7251b2e6a1775f2d5d81a73c0d376283c012cb500c6bfd93d0
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

Release files / astraea_adk-1.5.0-cp310-cp310-macosx_10_9_universal2.whl

Download URL astraea_adk-1.5.0-cp310-cp310-macosx_10_9_universal2.whl
Size 2.9 MB
Tags CPython 3.10 macOS 10.9+ universal2 (ARM64, x86-64)
SHA-256 checksum
How to use checksums
77ee010d97a0aa6ec574c880d379f4195e182d32d685eb3f3f672a7a113c2731
BLAKE2b-256 checksum
How to use checksums
f79c98e5583f0224a1c8884e0e7571d01d2060c4f4b032354cdee38a52f3c3a9
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

Security for astraea-adk 1.5.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

All 18 files in this release have been uploaded with provenance metadata.

Select from 18 files

Attestation for astraea_adk-1.5.0-cp315-cp315-win_amd64.whl

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 2, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page