text stringlengths 185 73.3k | repo stringlengths 7 100 | path stringlengths 4 146 | language stringclasses 7
values | hash stringlengths 16 16 | score float64 7 8.5 | stars int64 0 237k |
|---|---|---|---|---|---|---|
"""Normalized, provider-agnostic types.
These two structs are the *currency* of anypick: every obtainer translates its
raw payload into them, and the filter/pick layers operate on them and nothing
else. See docs/architecture.md.
"""
from __future__ import annotations
from dataclasses import dataclass, field
from typ... | gi-dellav/anypick | anypick-python/anypick/model.py | .py | 6ffad9dcc7ab7fb7 | 7.24 | 2 |
"""Vercel AI Gateway provider implementation (models only).
See docs/providers/vercel.md for the endpoint contract, response shape, and
mapping rules.
* :class:`VercelModelObtainer` — ``GET /v1/models`` (public; Bearer key optional).
The Vercel AI Gateway exposes **no benchmark feed**, so there is no
``VercelBenchma... | gi-dellav/anypick | anypick-python/anypick/vercel.py | .py | 4f21fa15ac591be4 | 7.24 | 2 |
#!/usr/bin/env python
# -*- coding: utf-8 -*-
"""
Agent Troubleshooter - 诊断工具
使用方法:
python diagnose.py "Codex 报 401 错误"
python diagnose.py "Claude 429 rate limit"
"""
import json
import sys
from pathlib import Path
# 确保以任意工作目录运行时都能找到同目录下的 utils 包
sys.path.insert(0, str(Path(__file__).resolve().parent))
# 导入... | XCZ-Huazhou/agent-troubleshoot-skill | diagnose.py | .py | 92ebabbba03bfbe7 | 7 | 0 |
"""
Agent Troubleshooter - 语义匹配和去重工具
功能：
1. 症状关键词提取和标准化
2. Jaccard 相似度计算
3. 框架名识别
4. 案例去重判断
"""
import json
import re
import sys
from datetime import date
from typing import List, Dict, Any, Tuple, Optional
from pathlib import Path
# 错误码标准化映射
ERROR_CODE_NORMALIZATION = {
# 认证错误
"401": "AUTH_ERROR",
"403"... | XCZ-Huazhou/agent-troubleshoot-skill | utils/matcher.py | .py | 051b2223814ebb77 | 7 | 0 |
#!/usr/bin/env python
# -*- coding: utf-8 -*-
""" batch-generate and tag wavetables, destination paths are hard-coded """
import os
import shutil
import wtcurve
import wtfile
import wttag
# destination paths for tagged files
wav_path = '/home/ftp/audio/Wavetables/WT1/My'
wt_path = os.path.expanduser('~/Music/Bitwig/... | porzione/wtcurve | gen_n_tag.py | .py | 18cd366e6ee06703 | 7.45 | 7 |
#!/usr/bin/env python
# -*- coding: utf-8 -*-
""" process wavetable files """
import sys
import os
import struct
import numpy as np
import soundfile as sf
WAV_SAMPLE_RATE = 44100
H2P_HEADER = """\
#defaults=no
#cm=OSC
Wave=2
<?
float Wave[%d];
"""
H2P_FMULT = 0.999969
# u-he Zebra 2 OSC format is fixed
H2P_NUM_WAV... | porzione/wtcurve | wtfile.py | .py | ef4e3f32b9a3fb38 | 7.45 | 7 |
"""
Agent 2 — Clause Miner
Uses Gemini (google-genai) to read raw contract text and extract every
AI-likeness / voice-clone / digital-double consent clause into a structured,
citable form. This is the first point where the Google Cloud AI SDK is
imported and called at runtime, per the hackathon's technical requirement... | asmita-ai/consent-continuity-agent | agents/clause_miner.py | .py | 4627bd4db7e5977e | 7 | 0 |
"""
Agent 1 — Ingestor
Normalizes heterogeneous inputs (contract PDFs, call sheets, dubbing scripts,
promo/campaign plans, game tie-in briefs) into plain text ready for the
Clause Miner. In production this would use Vertex AI Document AI for PDFs;
for the hackathon build we support .txt/.pdf and a pluggable extractor.... | asmita-ai/consent-continuity-agent | agents/ingestor.py | .py | 7c7ed4adabe30157 | 7 | 0 |
"""
Core data models for the Continuity rights ledger.
These are the machine-checkable structures that Agent 3 (Policy Compiler)
produces from Agent 2's (Clause Miner) raw extraction, and that Agent 4
(Usage Monitor) queries against every planned-use event.
"""
from __future__ import annotations
from datetime import... | asmita-ai/consent-continuity-agent | models/schema.py | .py | 7d4277dbca21b760 | 7 | 0 |
"""Configuration management for Transcript Generator."""
import json
import os
import re
from dataclasses import asdict, dataclass, field
from pathlib import Path
from typing import Any, Dict, List, Optional, Union
# Valid export format options
VALID_FORMATS = {"text", "timestamped", "srt", "vtt", "json", "csv", "mar... | advaitpr/transcript-generator | src/transcript_generator/config.py | .py | 64ccb096ccfdaa6f | 7 | 0 |
"""URL parser and video ID extractor for video URLs."""
import re
from urllib.parse import parse_qs, urlparse
# Regular expression for a valid 11-character YouTube video ID
VIDEO_ID_REGEX = re.compile(r"^[a-zA-Z0-9_-]{11}$")
# Patterns for extracting video ID from various URL formats
URL_PATTERNS = [
# Standard ... | advaitpr/transcript-generator | src/transcript_generator/url_parser.py | .py | 84d7ff799ac49813 | 7 | 0 |
"""audio conversion helpers.
ffmpeg does the heavy lifting: anything it can decode gets turned into
Ogg Vorbis at the pen's native settings (mono 22050 Hz).
ffmpeg is found via the RAV_FFMPEG env var, then PATH, then the static
binary that ships with imageio-ffmpeg (if installed).
"""
import os
import re
import shut... | Markthegamer108/RAVage | rav_tool/convert.py | .py | 5fd40ffd60d0c63c | 7.24 | 2 |
"""任务管理器：子进程隔离跑翻译，UI 崩了不影响翻译。
每个任务 = 独立子进程（python -m translator.cli 变体），通过
JSONL 事件流文件 + 控制管道与父进程通信。单任务模型：
同一时刻只允许一个翻译在跑（LLM 并发/缓存 DB 都是单写者设计）。
"""
from __future__ import annotations
import json
import os
import subprocess
import sys
import threading
import time
import uuid
from pathlib import Path
from translator.c... | ShZbz/pdf-translator | server/jobs.py | .py | 9fee7d77dd6a8395 | 7.15 | 1 |
"""P2 验收单测（SCHEME §6 P2）：
- batch 协议 roundtrip（LLM 返回 JSON → 译文落位）
- 缓存二次运行 0 调用
- max_llm_calls 触顶行为（不崩、保留原文、stderr 警告）
- [FORMULA_n] 计数守恒
- 坏响应重试一次后降级保留原文
"""
from __future__ import annotations
import json
import threading
import time
import sys
from pathlib import Path
from types import SimpleNamespace
sys.path.in... | ShZbz/pdf-translator | tests/test_smoke.py | .py | cab2744f7a301985 | 7.65 | 1 |
"""v0.4.0 UI 集成单测：事件流 / 暂停恢复 / 取消 / 锁安全。
全部零网络零 API key（FakeLLM 复用 test_smoke 的模式）。
"""
from __future__ import annotations
import json
import sys
import threading
import time
from pathlib import Path
from types import SimpleNamespace
sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
from translator.contr... | ShZbz/pdf-translator | tests/test_v040.py | .py | dc8c767f28102b6d | 7.65 | 1 |
"""协作式任务控制：暂停/恢复/取消（v0.4.0 UI 支持）。
语义（批间暂停）：
- 检查点粒度 = 页级（布局/渲染循环）+ 批级（LLM 调用前）。
- pause 后正在飞行中的那一次 LLM 请求会跑完才停，不半路掐断
（省 token、防半截缓存）。实际延迟几秒内。
- cancel 在下一个检查点立即生效，抛 JobCancelled 向上传播；
调用方负责收尾（管线内已保证不落半成品 PDF：先写 .tmp 再 rename）。
并发纪律：所有公开方法遵循「锁内改状态、锁外发通知」——
notify 必须在释放 self._lock 之后执行（非重入锁，锁内 notify = 死锁）。
"""
fr... | ShZbz/pdf-translator | translator/control.py | .py | a598c914a7ab3b74 | 7.15 | 1 |
"""进度事件流（v0.4.0 UI 支持）。
设计：
- 事件是纯 dict（JSON 可序列化），经回调推给消费者（server 端转 JSONL/轮询快照）。
- 回调永远不抛异常出去——UI 故障不能拖死翻译管线。
- CLI 不装回调时零开销（一次 None 判断）。
"""
from __future__ import annotations
import threading
import time
from typing import Callable
class EventSink:
"""线程安全的事件收集器。
on_event: callable(dict) -> None，在调用者线程... | ShZbz/pdf-translator | translator/events.py | .py | fe236ae250150db7 | 7.15 | 1 |
"""文字层提取：page → 结构化块列表（BBox/size/flags/font）。
<50 字符的扫描页候选走 OCR 降级（D7），OCR 引擎为可选依赖，
未安装时该页保留空文本并计入警告。
"""
from __future__ import annotations
import pymupdf
def page_has_text_layer(page, min_chars: int = 50) -> bool:
"""D7: 文字层字符数 < min_chars 视为扫描页候选。"""
return len(page.get_text().strip()) >= min_chars
def... | ShZbz/pdf-translator | translator/extract.py | .py | dbc336a63e62687d | 7.15 | 1 |
"""D4 术语表：外部 YAML {src: dst}，注入 system prompt + 译后校验。"""
from __future__ import annotations
from pathlib import Path
import yaml
class Glossary:
def __init__(self, mapping: dict[str, str] | None = None):
self.mapping: dict[str, str] = dict(mapping or {})
@classmethod
def load(cls, path: str | P... | ShZbz/pdf-translator | translator/glossary.py | .py | 08f4da0e360bf6f7 | 7.15 | 1 |
"""P4: preprocess.py — D5 水印两层策略(文字层 wmremover 式清理)。
三层递进(SCHEME D5,复用 wmremover.py 逻辑 + P4 实测修复):
1. 独立短流(<500B)含水印关键词 → 清空
2. 长流/Form XObject 中斜切/旋转矩阵 Tm + 水印关键词的 BT..ET 块 → 移除块
(P4 实测:水印常藏在 Form XObject 里;斜切矩阵形如 "1 0 0.21256 1",
原"四小数"正则漏检 → 放宽为允许 b/c 为 0)
3. Watermark/Stamp 注释 → 删除
cv2.inpaint 仅用于... | ShZbz/pdf-translator | translator/preprocess.py | .py | e5df1e87b4cb0dd3 | 7.15 | 1 |
"""修复 _wrap_cjk: Latin 词不拆行(词边界断行), CJK 保持逐字断行。
v0.2.2 任务1/5 排版包:英文保留段/双语原文层被逐字拆碎的根因是
贪心逐字断行对 Latin 文本无词边界概念。改为混合策略:
- 连续 Latin/digit 串视为不可分 token
- 行首放不下整个 token 时整词压到下一行;超长 token(>行宽)硬切兜底
- CJK 字符维持逐字+避头尾
"""
from __future__ import annotations
import re
import pymupdf
# 行首禁则标点（不可出现在行首，悬挂到上一行行尾）
_NO_LINE_START = s... | ShZbz/pdf-translator | translator/wrap_mixed.py | .py | ef26f1640ee9f16f | 7.15 | 1 |
#!/usr/bin/env python3
"""activity-emitter — async Claude Code hook that journals the core's activity
as AWP activity objects (Activity outbox Phase 2, step 1 — owner's pick
2026-07-24, delivered via the human-action bridge's first live card).
The AWP roadmap's Phase 2 is a durable Agent Activity outbox: the owner (an... | sonichi/sutando | hooks/activity-emitter.py | .py | 8a417efc6dd3f13b | 8 | 387 |
#!/usr/bin/env python3
"""gmail-write-guard — PreToolUse hook that denies the claude.ai Gmail MCP
connector's WRITE-scoped tools and routes writes to the IMAP/SMTP path.
Why (field report 05cb849a, michael@actoneventures.com, 2026-07-13): every
Gmail WRITE operation through the claude.ai connector is unreliable or bro... | sonichi/sutando | hooks/gmail-write-guard.py | .py | a0dc70dc7b2c00e9 | 8 | 387 |
#!/usr/bin/env python3
"""result-file-marker-guard — PreToolUse hook that DENIES writing a result body
whose ``[file:|send:|attach:]`` marker points outside the send allowlist **for
the adapter that will actually deliver it**.
Why (owner incident 2026-08-04, #susan): the agent finished a 6-minute video,
wrote ``[file:... | sonichi/sutando | hooks/result-file-marker-guard.py | .py | e81bd69f743cea53 | 8 | 387 |
#!/usr/bin/env python3
"""Skill-usage telemetry — PostToolUse[Skill] hook.
Emits ONE anonymous ``feature_used {feature: "skill:<name>"}`` product-telemetry
event every time the core invokes a skill (the `Skill` tool). This is the
chokepoint that broadens feature-usage coverage from the two hand-instrumented
scripts (m... | sonichi/sutando | hooks/skill-usage-telemetry.py | .py | e1ea01ff51c2b587 | 8 | 387 |
#!/usr/bin/env python3
"""skip-ask-user-question — PreToolUse hook that blocks the interactive
`AskUserQuestion` tool in Sutando's headless core session.
Why: the core agent runs NON-INTERACTIVELY — src/agent/claude/cli/start-cli.sh
launches it with `--dangerously-skip-permissions` inside a tmux pane, driven
over `--r... | sonichi/sutando | hooks/skip-ask-user-question.py | .py | 13665d1dfc5f5fa1 | 7 | 387 |
"""Fail-closed secret redaction for persisted inbound chat content.
The explicit ``vault set KEY VALUE`` path stores named values in Keychain.
This module covers the other common case: somebody pastes a token into ordinary
Discord/Slack prose. Known values are replaced before task files, owner-activity
state, prompt s... | sonichi/sutando | packages/ag2-sparrow/ag2_sparrow/chat_secret_filter.py | .py | cc059b9c06981d84 | 8 | 387 |
#!/usr/bin/env python3
"""Recovery for a `[deduped: <holder>]` result whose holder never answered.
`result_markers.dedup_decision` decides; this binds that decision to a
workspace and performs the filesystem half, so every adapter keeps only its
own routing and notification.
Returns a `(action, payload)` plan rather ... | sonichi/sutando | packages/ag2-sparrow/ag2_sparrow/dedup_recovery.py | .py | 194467cbc6f30ee2 | 8 | 387 |
"""Delivery Core contract: the channel-neutral seam between local ownership
(ClaimBackend), the external side effect (DeliveryProvider), and the drain
loop (DeliveryCore in core.py).
Identity model (three types, never conflated):
- item_id: stable logical message identity, assigned at publish.
- ClaimToken: one local-... | sonichi/sutando | packages/ag2-sparrow/ag2_sparrow/delivery_core/contract.py | .py | f7b11971da558165 | 8 | 387 |
"""DeliveryCore: the drain loop. The ONLY component that reads
DeliveryOutcome semantics and ProviderCapabilities — channel code never
decides retry/UNKNOWN rules (acceptance criterion 2).
Retry policy (normative):
- only a confirmed NOT_DELIVERED auto-retries;
- OUTCOME_UNKNOWN parks unless capabilities license recon... | sonichi/sutando | packages/ag2-sparrow/ag2_sparrow/delivery_core/core.py | .py | 68f79fc9a6ad800a | 8 | 387 |
"""Migration fencing: single-protocol-per-epoch (seam doc §4).
One logical item is interpreted by exactly ONE claim protocol within an
epoch. Migration = lock out drainers -> one-shot convert (each item
individually atomic) -> write the version fence -> start only the new
drainer. The fence is written LAST: a crash an... | sonichi/sutando | packages/ag2-sparrow/ag2_sparrow/delivery_core/migration.py | .py | 7ab8f786712bf36a | 8 | 387 |
"""event_channel — Sparrow's persistent Workspace-Event delivery channel (#AWP P0).
An ADDITIVE, ISOLATED channel that runs ALONGSIDE task delivery and never touches
it. It keeps an outbound SSE connection to `/v1/events/stream`, writes every
authorized event durably to the local EventInbox (at-least-once), and resume... | sonichi/sutando | packages/ag2-sparrow/ag2_sparrow/event_channel.py | .py | fc8dc90164dee131 | 8 | 387 |
"""Restart recovery for proactively delivered result files.
Bridges claim ``proactive-*.txt`` as ``.sending``; this restores claims a crash
stranded so every adapter applies one collision and failure policy at startup.
"""
from __future__ import annotations
import os
import re
from pathlib import Path
from typing im... | sonichi/sutando | packages/ag2-sparrow/ag2_sparrow/proactive_recovery.py | .py | c489835ce1ef1c47 | 8 | 387 |
#!/usr/bin/env python3
"""Readiness of a `results/<task-id>.txt` file, for every delivery consumer.
The single owner of "is this result file ready to send?". Adapters bind their
own resolved results directory and keep only provider-specific delivery; they
must not re-implement the check.
A result path can exist befor... | sonichi/sutando | packages/ag2-sparrow/ag2_sparrow/result_ready.py | .py | dc48c98ae0747325 | 8 | 387 |
"""Shared file-attachment allowlist for `[file:|send:|attach:]` markers.
Single source of truth for the policy that decides whether an agent-
emitted file marker can be delivered to the owner's Discord DM /
channel. Used by:
- ``src/discord-bridge.py`` — live WS-connected bridge
(``discord.File(path)``).
- ``... | sonichi/sutando | packages/ag2-sparrow/ag2_sparrow/send_allowlist.py | .py | b416718b64facbfb | 8 | 387 |
"""Classify an outbound-send failure as transient (retry) or permanent (park).
Parking is the safe default: only a KNOWN transient retries.
"""
from __future__ import annotations
from pathlib import Path
# 4xx timing cases only. 5xx is handled as a RANGE below, because enumerating it
# silently parked the Cloudflar... | sonichi/sutando | packages/ag2-sparrow/ag2_sparrow/send_failure_policy.py | .py | 944eb3c5610d186e | 8 | 387 |
#!/usr/bin/env python3
"""The Team-tier guardrail prose, shared by every surface that admits Team work.
Two consumers interpret the same policy — the workstream session worker (which
used to wrap it around a spawned Team session) and the AG2 Space gateway (which
must now emit it in-band, because closing the Team sessi... | sonichi/sutando | packages/ag2-sparrow/ag2_sparrow/team_guardrail.py | .py | 9837295d360c2b03 | 8 | 387 |
"""Pure, dependency-free `vault set KEY VALUE` grammar — regex + redact-only.
Canonical source for `vault_intercept.py` and the vendored ag2-sparrow copy; no imports beyond `re`.
"""
from __future__ import annotations
import re
# Separator is whitespace or `=`; KEY stops at the first space/`=` so `=` isn't swallowed... | sonichi/sutando | packages/ag2-sparrow/ag2_sparrow/vault_set_grammar.py | .py | c3aa73e81c0e0f5c | 7 | 387 |
"""_post_task_ack — self-healing retry when the gateway lacks /v1/tasks/<id>/ack.
The worker acks each pulled task so the broker can surface a "received" state.
When the broker lacks the endpoint it 404s; the worker must back off — but NOT
permanently, or a broker that later *deploys* the endpoint is never picked up
u... | sonichi/sutando | packages/ag2-sparrow/tests/test_ack_retry.py | .py | d46f085c12c4ed1f | 8.5 | 387 |
"""Bounded DNS resolution — a hung resolver must raise (so the poll loop can
emit "reconnecting" and retry) instead of wedging the process forever.
Regression guard for the 2026-07-25 tester incident: the gateway sat in a
"reconnecting" state indefinitely because getaddrinfo (which has no native
timeout) blocked the l... | sonichi/sutando | packages/ag2-sparrow/tests/test_dns_timeout.py | .py | a1727d41b910257f | 8.5 | 387 |
"""The env tier must strip quotes, like the three file tiers already do.
A user who writes the credential quoted —
REMOTE_TASK_TOKEN='https://gw.example/relay|<secret>'
— gets a working bridge when the value is read from the FILE (three readers
call `.strip().strip("'\\"")`) and a permanently unauthorized one wh... | sonichi/sutando | packages/ag2-sparrow/tests/test_env_token_quotes.py | .py | f66449501964d261 | 8.5 | 387 |
"""Tests for event_consumer (AWP P1) — inbox → taskify → tasks/. Covers ambient
trust boundary, held-events-not-consumed (no loss), idempotent re-drain, and
skip-settles-immediately. Self-contained; exit 0/1."""
import os
import sys
import tempfile
from pathlib import Path
sys.path.insert(0, str(Path(__file__).resolve... | sonichi/sutando | packages/ag2-sparrow/tests/test_event_consumer.py | .py | 1ce87de2d8e59f8a | 8.5 | 387 |
"""Tests for the P0 event delivery channel — durable inbox + persistent SSE
consumer. Covers the friend's fault-recovery acceptance criteria: at-least-once
dedup, cursor recovery, crash-before-durable safety, channel isolation, fatal
auth stop. Self-contained (stdlib + a mocked urlopen). Exit 0/1."""
import io
import o... | sonichi/sutando | packages/ag2-sparrow/tests/test_event_inbox_channel.py | .py | 0778635d970427c3 | 8.5 | 387 |
"""_emit_gateway_status — connection-liveness sidecar (state/gateway-status.json).
Covers the shape a local supervisor reads and the last_ok_ts preservation that
lets it show "last connected N s ago" while reconnecting.
"""
import json
import os
import tempfile
import importlib
import sys
import pathlib
def _load(tmp)... | sonichi/sutando | packages/ag2-sparrow/tests/test_gateway_status.py | .py | 2edc623ac0cda31c | 8.5 | 387 |
"""in-flight ledger — crash-before-ack recovery (v1 freeze gate #142 item 6).
A task pulled from the broker but not yet completed (`POST /v1/results`) lives in
the persisted in-flight set. If the worker restarts between pull and result-POST,
the set must survive so the result drain re-acks it — otherwise the reply is ... | sonichi/sutando | packages/ag2-sparrow/tests/test_inflight_recovery.py | .py | f3f78aa4e0eff8ee | 8.5 | 387 |
"""FastAPI backend entry point for the Insurance Hybrid RAG application."""
from __future__ import annotations
import logging
from contextlib import asynccontextmanager
from typing import Annotated
import uvicorn
from fastapi import FastAPI, Header, HTTPException
from fastapi.middleware.cors import CORSMiddleware
fro... | Rabilkhan786/health-agentic-rag | app.py | .py | a5e31c9f89faf894 | 7 | 0 |
"""Single validated configuration entry point for the application."""
from __future__ import annotations
import os
from dataclasses import dataclass
from pathlib import Path
from typing import Any
import yaml
from dotenv import load_dotenv
REQUIRED_SECTIONS = (
"project",
"models",
"pinecone",
"retr... | Rabilkhan786/health-agentic-rag | config/settings.py | .py | 3ce0f10030d92125 | 7 | 0 |
"""Run the RAGAS evaluation suite against evaluation/rag_dataset.json.
Usage: uv run python evaluation/run_ragas.py
uv run python evaluation/run_ragas.py --smoke (1 per topic)
WHAT THIS MEASURES, AND WHAT IT DOES NOT: this scores the retrieval half of
the copilot -- can it find the clause that answers a qu... | Rabilkhan786/health-agentic-rag | evaluation/run_ragas.py | .py | ba95af709e17392c | 7 | 0 |
"""Root entry point for the PDF-to-Pinecone indexing workflow.
Runs the section-aware pipeline (src/ingestion/pipeline.py) over every PDF
in Data/insurance_documents/ and upserts the result into Pinecone.
Upserts are idempotent -- document IDs hash the chunk's own content -- so
re-running overwrites the same vectors r... | Rabilkhan786/health-agentic-rag | main.py | .py | cf7379820c5a80ab | 7 | 0 |
"""Layer 1 — the tool-calling agent, built by LangChain's create_agent.
create_agent owns the whole reason/act loop, parallel tool execution, and
session memory. Nothing in this file hand-rolls any of that.
Context.customer_id is the fix for the old "please give me your policy ID"
bug: it is passed at invoke time and... | Rabilkhan786/health-agentic-rag | src/agent/agent.py | .py | 5a0f4856cbb356de | 7 | 0 |
"""The typed recommendation an employee reviews.
WHY the numbers are assembled here rather than asked of the model: the
deterministic engine has already decided the status and computed the payable
amount. Handing those to the LLM and asking for them back is a round trip
that can only lose -- it once returned a payable... | Rabilkhan786/health-agentic-rag | src/agent/recommendation.py | .py | 056803903b04bbe5 | 7 | 0 |
"""Layer 2 — the outer workflow: a claim goes in, a reviewed decision comes out.
claim submitted -> eligibility -> agent -> review (pauses) -> persist -> END
chat question -> agent -> END
The compiled create_agent graph from Layer 1 is added here as an ordinary node
(subgraph-as-node), so the claim path end... | Rabilkhan786/health-agentic-rag | src/agent/workflow.py | .py | 8d3891ca6b466b3c | 7 | 0 |
"""In-process TTL cache for hybrid retrieval results.
WHY: a full retrieval is dense search + sparse search + RRF + cross-encoder
rerank, which costs roughly 4 seconds. Demo users ask the same handful of
questions repeatedly, and the eligibility engine re-queries the same treatment
for coverage and exclusions, so the ... | Rabilkhan786/health-agentic-rag | src/cache/rag_cache.py | .py | 720c6861bcd42901 | 7 | 0 |
"""SQLite-backed audit trail: what the AI recommended vs what the employee did."""
from __future__ import annotations
import json
import logging
import sqlite3
from uuid import uuid4
from src.utils.sqlite_store import SqliteStore
from .models import SCHEMA
logger = logging.getLogger(__name__)
# The three calls the... | Rabilkhan786/health-agentic-rag | src/decisions/store.py | .py | 3c42134ebf8b9257 | 7 | 0 |
"""How the engine represents a policy fact it may or may not actually know.
WHY this exists: the engine used to return plain numbers, so "this policy
states no co-payment" and "nobody could find out whether this policy states a
co-payment" arrived at the calculator as the same value -- None, read as 0%.
The claim was ... | Rabilkhan786/health-agentic-rag | src/eligibility/facts.py | .py | e21792ee7eff6398 | 7 | 0 |
"""Pull numbers out of policy clause text.
WHY this exists: structured facts (sub-limits, waiting periods, co-pays,
deductibles) are only in SQL for policies someone has already curated. A
customer arriving with any of the other indexed policies has no rows at all,
and the engine used to treat that as "no limit applie... | Rabilkhan786/health-agentic-rag | src/eligibility/parsing.py | .py | 4c5c12436807945a | 7 | 0 |
"""Separate a PDF page's prose from its tables.
WHY this file exists: PyMuPDF's ``page.get_text()`` dumps table cell content
straight into the text stream. That means a sliding-window chunker happily
produces chunks like "Cataract 25000 Hernia 30000 waiting period 24" -- cell
values glued to unrelated prose, which ret... | Rabilkhan786/health-agentic-rag | src/ingestion/page_parser.py | .py | 38a1dd364684e79e | 7 | 0 |
"""Re-index every policy PDF: clean text to chunks, tables to their homes.
WHY: this replaces the single-path Unstructured loader for the policy
corpus. Each page is split into prose and tables first, so a chunk never
contains stray table cells and a table never gets chopped into chunks.
"""
from __future__ import ann... | Rabilkhan786/health-agentic-rag | src/ingestion/pipeline.py | .py | 75f2a64ff87431ae | 7 | 0 |
"""Turn classified table rows into natural-language sentences for Pinecone.
WHY: an embedding model cannot make sense of a bare row like
``["Cataract", "Rs 25,000"]``. Rewritten as a full sentence that names the
insurer, the UIN and the page, the same row becomes a retrievable fact with
a citation baked in.
"""
from _... | Rabilkhan786/health-agentic-rag | src/ingestion/table_converter.py | .py | 333310a8bae4b983 | 7 | 0 |
"""Hybrid retrieval, composed from LangChain retrievers instead of by hand.
The shape is the one the reference book describes -- run a dense and a sparse
retriever in parallel, fuse their rankings, then rerank the survivors with a
cross-encoder -- but built with the current LangChain 1.x classes:
EnsembleRetrieve... | Rabilkhan786/health-agentic-rag | src/retrieval/retrievers.py | .py | e84cc071d69bbde9 | 7 | 0 |
"""Agent-facing tools over the CRM store: customers, policies, claims.
WHY these tools take no customer_id: it is injected at invoke time via
ToolRuntime.context, so it never appears in the schema the LLM sees. The
model therefore cannot reach another customer's records, and cannot ask for
an ID the employee already e... | Rabilkhan786/health-agentic-rag | src/tools/crm_tools.py | .py | 2760bc61be3e0867 | 7 | 0 |
"""Agent-facing tools over the hybrid RAG pipeline.
Every function narrows retrieval with a Pinecone metadata filter (topic
and/or policy UIN) instead of searching the whole index — a question about
waiting periods should only compete against waiting_period chunks.
The filter matches on `topics`, the list of every to... | Rabilkhan786/health-agentic-rag | src/tools/rag_tools.py | .py | 17f9fe547d50f7e8 | 7 | 0 |
"""Shared SQLite plumbing for the three stores in this project.
CRMStore, PolicyDataStore and DecisionStore each opened their own connection,
set the same row factory, and ran their own schema on first use -- the same
fifteen lines written three times. They differ only in which schema they
create and what they log, so... | Rabilkhan786/health-agentic-rag | src/utils/sqlite_store.py | .py | b8148af89e694cd0 | 7 | 0 |
"""Adapters for the existing dense and hosted sparse Pinecone indexes.
WHY this is not langchain-pinecone's PineconeVectorStore: that class covers
one dense index, and this project runs two indexes of different kinds. The
sparse half is a Pinecone *hosted* sparse index -- created with
create_index_for_model() and quer... | Rabilkhan786/health-agentic-rag | src/vectorstores/pinecone_store.py | .py | da697b67b52ff402 | 7 | 0 |
"""Claims Copilot - the employee-facing frontend.
Run with: uv run streamlit run streamlit_app.py
This is an INTERNAL tool. The person using it is a claims employee, not a
customer. The screen shows a recommendation and the employee decides: the
Approve / Edit / Reject actions at the bottom are the point of the whol... | Rabilkhan786/health-agentic-rag | streamlit_app.py | .py | 2c6df0b2c09496e7 | 7 | 0 |
"""Deterministic cache-key derivation.
``pickle.dumps`` is not a safe basis for cache keys: the byte stream for a
``set`` (or any object reducing to one) depends on string hash randomisation,
so the same logical key hashes differently in every process. This module walks
the value instead and emits a canonical encoding... | farfarfun/farcache | src/farcache/_keys.py | .py | 1d92937d7399dab7 | 7 | 0 |
"""Internal utilities shared across farcache modules."""
from __future__ import annotations
import inspect
from collections.abc import Callable
from typing import Any
__all__ = ["bind_args", "namespace_of"]
def bind_args(
signature: inspect.Signature,
args: tuple[Any, ...],
kwargs: dict[str, Any],
) ->... | farfarfun/farcache | src/farcache/_utils.py | .py | 99e9f7e4000c1f61 | 7 | 0 |
"""In-memory caches, backed by :mod:`cachebox` eviction policies.
Every decorator here works both bare and called::
@lru_cache
def f(x): ...
@lru_cache(maxsize=500)
def g(x): ...
The wrapper exposes the underlying policy as ``f.cache``, so ``f.cache.clear()``
empties it and ``len(f.cache)`` reports ... | farfarfun/farcache | src/farcache/box.py | .py | dc4e4eedbb09bde7 | 7 | 0 |
"""SQLite-backed function cache, built on :mod:`diskcache`."""
from __future__ import annotations
import os
from collections.abc import Callable, Iterable
from hashlib import sha256
from typing import Any
from diskcache import Cache
from ._base import MISSING, CacheStore, FunctionCache
from ._utils import namespace... | farfarfun/farcache | src/farcache/disk.py | .py | 7d41cfdbec03e2f4 | 7 | 0 |
#!/usr/bin/env python3
"""Two promises of install.py that failed silently once, so they get a check each.
⛔ WHY THESE TWO AND NOTHING ELSE HERE. Both were the same shape of bug: the script SAID
the right thing and did the wrong thing.
1. `--check` was honoured by the statusline half only, so `--all --check` printed ... | Dino9021/dispatch-guard | test_install.py | .py | dd7ef68050b75369 | 7.5 | 0 |
"""AWS client wrapper for the AWS Infrastructure integration."""
from __future__ import annotations
import logging
import boto3
from botocore.config import Config
_LOGGER = logging.getLogger(__name__)
# Apply a connect and read timeout to all boto3 clients.
# Without this, a hung or slow AWS endpoint can block a co... | ianpleasance/home-assistant-aws-infrastructure | custom_components/aws_infrastructure/aws_client.py | .py | b80a77f21df7cc36 | 7 | 0 |
"""Config flow for AWS Infrastructure integration."""
from __future__ import annotations
import json
import logging
from typing import Any
import voluptuous as vol
from homeassistant import config_entries
from homeassistant.core import callback
import homeassistant.helpers.config_validation as cv
from .const import... | ianpleasance/home-assistant-aws-infrastructure | custom_components/aws_infrastructure/config_flow.py | .py | 22abcc694222cfa4 | 7 | 0 |
import logging
from datetime import timedelta
from django.db.models import Count, OuterRef, Subquery
from django.utils import timezone
from rest_framework import generics, permissions, serializers
from rest_framework.response import Response
from rest_framework.views import APIView
from drf_spectacular.utils import ex... | coneshare/coneshare | backend/analytics/views.py | .py | 952a36892c43e463 | 7.78 | 35 |
from collections import defaultdict
from datetime import timedelta
import logging
from zoneinfo import ZoneInfo
from django.utils.dateparse import parse_datetime
from django.utils import timezone
from django.utils.translation import gettext as _, override as translation_override
from django.core.mail import send_mail
... | coneshare/coneshare | backend/automations/emails.py | .py | 1bb5e1e706fc5c0a | 7.78 | 35 |
import logging
import uuid
from core.models import Organization, User
from .models import AutomationDelivery, AutomationRule, AutomationDestination
from .constants import EMAIL_COALESCE_DEBOUNCE_SECONDS
logger = logging.getLogger(__name__)
def _rule_matches_scope(rule, payload):
share_link_id = payload.get('sh... | coneshare/coneshare | backend/automations/services.py | .py | 78692ad66830b017 | 7.78 | 35 |
'''
pip install flask
export CONESHARE_SIGNING_SECRET='supersecret' # optional for local testing
python coneshare_webhook_receiver.py
'''
import hashlib
import hmac
import json
import os
from flask import Flask, request, jsonify
app = Flask(__name__)
# Set this to the same value as AutomationDestination.signing_s... | coneshare/coneshare | backend/automations/webhook_receiver.example.py | .py | 9bf2dc1530b17d4b | 7.78 | 35 |
import os
import re
def get_client_ip(request):
"""
Gets the real client IP address from a request, prioritizing a trusted
X-Real-IP header set by a proxy. If not present, it falls back to
X-Forwarded-For and then REMOTE_ADDR.
"""
real_ip = request.META.get('HTTP_X_REAL_IP')
if real_ip:
... | coneshare/coneshare | backend/backend/utils.py | .py | dfc53ad8a86d8f15 | 7.78 | 35 |
import os
import pytest
from django.contrib.auth import get_user_model
from rest_framework.test import APIClient
from compile_po import BASE_LOCALE, make_mo
from core.models import Organization
User = get_user_model()
DEFAULT_TEST_PASSWORD = "StrongPassword123!"
@pytest.fixture(autouse=True, scope="session")
def e... | coneshare/coneshare | backend/bdd/conftest.py | .py | cc047b51a3b0a02d | 8.28 | 35 |
from pytest_bdd import given
from rest_framework import status
@given("I am an authenticated user", target_fixture="user_context")
def user_context(user, api_client):
"""The user and api_client fixtures handle authentication."""
return {"user": user, "api_client": api_client}
@given("my document list is emp... | coneshare/coneshare | backend/bdd/step_definitions/common_steps.py | .py | f69675d0c057a5cc | 7.78 | 35 |
import pytest
from unittest.mock import patch, MagicMock
from django.core.files.uploadedfile import SimpleUploadedFile
from pytest_bdd import parsers, scenario, then, when
from rest_framework import status
pytest_plugins = "bdd.step_definitions.common_steps"
@pytest.mark.django_db
@scenario('../features/document_upl... | coneshare/coneshare | backend/bdd/step_definitions/test_document_upload.py | .py | 92c3cac713055dab | 8.28 | 35 |
import pytest
from unittest.mock import patch, MagicMock
from django.core.files.uploadedfile import SimpleUploadedFile
from pytest_bdd import parsers, scenario, given, when, then
from rest_framework import status
from documents.models import Document, DocumentVersion
# Make common steps available
pytest_plugins = "bd... | coneshare/coneshare | backend/bdd/step_definitions/test_document_version_upload.py | .py | f658ff52fc154784 | 8.28 | 35 |
import pytest
from pytest_bdd import given, parsers, scenario, then, when
from rest_framework import status
from documents.models import Document
from sharelinks.models import ShareLink, ViewSession, Viewer
pytest_plugins = "bdd.step_definitions.common_steps"
@pytest.mark.django_db
@scenario('../features/share_link... | coneshare/coneshare | backend/bdd/step_definitions/test_share_link_analytics.py | .py | 8140b7795332003d | 8.28 | 35 |
import pytest
from pytest_bdd import parsers, scenario, given, when, then
from rest_framework import status
from documents.models import Document
from sharelinks.models import ShareLink
# Make common steps available
pytest_plugins = "bdd.step_definitions.common_steps"
@pytest.mark.django_db
@scenario(
'../featu... | coneshare/coneshare | backend/bdd/step_definitions/test_share_link_view.py | .py | 6e9a9cb105297ff7 | 8.28 | 35 |
import re
from urllib.parse import urlparse, parse_qs
import pytest
from django.contrib.auth import get_user_model
from django.core.cache import cache
from django.urls import reverse
from pytest_bdd import scenario, given, when, then, parsers
from rest_framework import status
from core.models import AppConfiguration
... | coneshare/coneshare | backend/bdd/step_definitions/test_user_authentication.py | .py | ce9f9445b21a8c66 | 8.28 | 35 |
import pytest
from unittest.mock import patch
from pytest_bdd import scenario, given, when, then, parsers
from documents.models import Document
from sharelinks.models import ShareLink, ViewSession, PageView
pytest_plugins = "bdd.step_definitions.common_steps"
@pytest.mark.django_db
@scenario('../features/view_track... | coneshare/coneshare | backend/bdd/step_definitions/test_view_tracking.py | .py | 940ce9304e709d66 | 8.28 | 35 |
"""CLI for Encrypted P2P Chat."""
import typer
import base64
from p2pchat.crypto import IdentityKeys
from p2pchat.protocol import ChatSession
cli = typer.Typer(name="p2pchat", help="Educational Encrypted P2P Chat")
@cli.command()
def generate_identity():
"""Generate a new identity key pair."""
... | OpKnock/encrypted-p2p-chat | src/p2pchat/cli.py | .py | 09ca71ef42a14101 | 7 | 0 |
"""Educational cryptographic primitives for P2P chat."""
import os
import base64
import json
from typing import Tuple, Dict, Optional
from dataclasses import dataclass
from cryptography.hazmat.primitives.asymmetric.x25519 import X25519PrivateKey, X25519PublicKey
from cryptography.hazmat.primitives.asymmetric... | OpKnock/encrypted-p2p-chat | src/p2pchat/crypto.py | .py | 28cc641699791182 | 7 | 0 |
# config.py
import logging
from llama_index.core import Settings
logger = logging.getLogger("campus_rag")
Settings.llm = None
Settings.embed_model = None
Settings.chunk_size = 1024
Settings.chunk_overlap = 50
_llm_initialized = False
_embed_initialized = False
def init_llm() -> bool:
"""Initialize LLM. Returns... | allowan/USTC-107-let-ai-think-of-one | campus_rag/config.py | .py | 4b85bc2945b8cbb0 | 7 | 0 |
#data_loader.py
from llama_index.core import Document
from llama_index.core.node_parser import SentenceSplitter
import os
def load_documents_from_files(directory: str) -> list:
"""读取目录下所有 .txt 文件，每个文件为一个 Document"""
documents = []
if not os.path.isdir(directory):
return documents
for filename i... | allowan/USTC-107-let-ai-think-of-one | campus_rag/data_loader.py | .py | 49c34c7006b85912 | 7 | 0 |
import logging
from pathlib import Path
from llama_index.core import Document
_base = Path(__file__).resolve().parent
from .index_manager import RAGSystem
logger = logging.getLogger("campus_rag.query")
_rag = None
_public_retriever = None
_user_retrievers: dict[str, object] = {}
def _reset():
"""重置所有缓存状态，下次调用... | allowan/USTC-107-let-ai-think-of-one | campus_rag/query.py | .py | 3157d16554fc349f | 7 | 0 |
import json
import os
from pathlib import Path
from langchain.chat_models import init_chat_model
_SETTINGS_PATH = Path(__file__).resolve().parent.parent / "settings.json"
def read_json() -> dict:
"""读取 settings.json 中的 env 配置（文件不存在时返回空字典）。"""
try:
with open(_SETTINGS_PATH, encoding="utf-8") as f:
... | allowan/USTC-107-let-ai-think-of-one | model/config.py | .py | e55a76fc1b1ab37a | 7 | 0 |
"""
Auth service: topic CRUD. Delegates to campus_rag.auth.
"""
import logging
logger = logging.getLogger("server")
class AuthService:
"""Thin wrapper around campus_rag.auth for topic management."""
@staticmethod
def list_topics(username: str) -> list:
from campus_rag import list_topics
... | allowan/USTC-107-let-ai-think-of-one | server/services/auth_service.py | .py | 0c8f31e7db0abb2f | 7 | 0 |
"""
RAG service: search, personal data CRUD.
Delegates to campus_rag.query and campus_rag.index_manager.
"""
import logging
from llama_index.core import Document
logger = logging.getLogger("server")
class RAGService:
"""Encapsulates RAG operations: search and personal data management."""
@staticmethod
... | allowan/USTC-107-let-ai-think-of-one | server/services/rag_service.py | .py | c1fb76ca2040c16f | 7 | 0 |
"""Local structured schedule storage."""
from __future__ import annotations
import json
import sqlite3
from datetime import datetime, timezone
from pathlib import Path
DB_PATH = Path(__file__).resolve().parents[2] / "schedule.db"
# USTC's standard period ranges. Imported files can provide exact times; these
# rang... | allowan/USTC-107-let-ai-think-of-one | server/services/schedule_service.py | .py | 381302af4cba4feb | 7 | 0 |
"""
Sync service: pulls public notices from sync_server, updates local ChromaDB.
"""
import json
import logging
from pathlib import Path
import httpx
logger = logging.getLogger("server")
ROOT = Path(__file__).resolve().parent.parent.parent
SYNC_STATE_PATH = ROOT / "data" / "sync_state.json"
# Default sync_server U... | allowan/USTC-107-let-ai-think-of-one | server/services/sync_service.py | .py | 9b15840dfeb16226 | 7 | 0 |
"""
Database: document storage + change log for versioned sync.
"""
import os
import sqlite3
import json
from pathlib import Path
from datetime import datetime
ROOT = Path(__file__).resolve().parent
DB_PATH = ROOT / "data" / "sync_server.db"
def _get_conn() -> sqlite3.Connection:
DB_PATH.parent.mkdir(parents=Tr... | allowan/USTC-107-let-ai-think-of-one | sync_server/database.py | .py | cc40ceeb57a7fc7e | 7 | 0 |
"""
Sync Server — 公共通知数据库同步服务端。
纯文档存储 + 版本管理，不依赖 LLM / ChromaDB / Ollama。
启动: python main.py (默认端口 8001)
"""
import sys
from pathlib import Path
# Allow running as python main.py from within sync_server/
sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
import logging
from fastapi import FastAPI
fro... | allowan/USTC-107-let-ai-think-of-one | sync_server/main.py | .py | d5992812d646ae26 | 7 | 0 |
"""
Refresh the free-proxy-list mirror from the ProxyScrape v4 public API.
Writes:
proxies/all/data.{txt,json,csv}
proxies/protocols/{http,https,socks4,socks5}/data.{txt,json,csv}
proxies/countries/{ISO}/data.{txt,json,csv} (ISO-3166 alpha-2, lowercased)
proxies/countries/{ISO}/{protocol}/data.... | ProxyScrape/free-proxy-list | scripts/update.py | .py | 84326d2fc5f52b51 | 8 | 120 |
"""Command-line interface for ofplang.schedule.
Thin presentation layer over the library. Subcommands:
ofp-schedule validate [--kind ...] [--format ...] <file>...
ofp-schedule schedule <workflow> --env <env> [--document <doc>] [-o <file>] [--format yaml|json]
ofp-schedule visualize <plan> [--view device|w... | ofplang/schedule | ofplang/schedule/cli.py | .py | f23a1dc20189a0ed | 7 | 0 |
"""Diagnostics and validation result.
Shared by both schema validators. A diagnostic pins a stable error code
(SPECIFICATIONS.md §10) to a source position; `severity` distinguishes hard
errors from warnings. Warnings never make a document invalid (§9), so
`ValidationResult.ok` looks only at errors.
"""
from __future_... | ofplang/schedule | ofplang/schedule/core/diagnostics.py | .py | 0205c42512428e53 | 7 | 0 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.