3a00ff4d33
- Complete Fastify gateway with 8-stage pipeline - Circuit breaker (opossum) per model tier - Rate limiting per caller - Ban list validation (EN/DE/auto-detected) - TIP validator (SFF-8024, part numbers, wavelengths) - Prometheus metrics - pg-boss async queue - PostgreSQL audit log + review queue - 9 prompt templates (TIP, LinkedIn, ShieldX) - Learning engine scaffolding - Auto-learning: ban-list, few-shot, routing, prompt optimizer
173 lines
6.3 KiB
YAML
173 lines
6.3 KiB
YAML
id: tip_datasheet_extract
|
|
version: "1.0.0"
|
|
task_type: tip_datasheet_extract
|
|
description: Extract complete transceiver specifications from PDF datasheets converted to Markdown by Docling, including min/typ/max electrical values
|
|
model_preference: qwen2.5:14b
|
|
model_minimum: qwen2.5:7b
|
|
temperature: 0.05
|
|
max_tokens: 3000
|
|
output_format: json
|
|
|
|
system_prompt: |
|
|
You are a precision data extraction specialist for the Transceiver Intelligence Platform (TIP).
|
|
You receive transceiver datasheet content that has been converted from PDF to Markdown by Docling.
|
|
Your task is to extract ALL technical specifications with maximum accuracy.
|
|
|
|
Return ONLY valid JSON with this structure:
|
|
{
|
|
"part_number": "string",
|
|
"vendor": "string",
|
|
"form_factor": "SFP|SFP+|SFP28|QSFP+|QSFP28|QSFP-DD|OSFP|CFP|CFP2|CFP4|XFP|other",
|
|
"data_rate_gbps": number,
|
|
"protocol": "Ethernet|Fibre Channel|SONET/SDH|OTN|other",
|
|
"wavelength_nm": number or null,
|
|
"wavelength_channels": [numbers] or null,
|
|
"reach_m": number or null,
|
|
"connector": "LC|SC|MPO|RJ45|other",
|
|
"fiber_type": "SMF|MMF|DAC|AOC|null",
|
|
"electrical": {
|
|
"tx_power_min_dbm": number or null,
|
|
"tx_power_typ_dbm": number or null,
|
|
"tx_power_max_dbm": number or null,
|
|
"rx_sensitivity_min_dbm": number or null,
|
|
"rx_sensitivity_typ_dbm": number or null,
|
|
"rx_sensitivity_max_dbm": number or null,
|
|
"extinction_ratio_min_db": number or null,
|
|
"oma_sensitivity_dbm": number or null,
|
|
"center_wavelength_min_nm": number or null,
|
|
"center_wavelength_max_nm": number or null
|
|
},
|
|
"supply_voltage_v": number or null,
|
|
"supply_current_ma_max": number or null,
|
|
"power_consumption_w_max": number or null,
|
|
"temperature_case_min_c": number or null,
|
|
"temperature_case_max_c": number or null,
|
|
"temperature_operating_min_c": number,
|
|
"temperature_operating_max_c": number,
|
|
"storage_temp_min_c": number or null,
|
|
"storage_temp_max_c": number or null,
|
|
"humidity_operating_pct_max": number or null,
|
|
"dom_support": true|false,
|
|
"standards_compliance": ["string"],
|
|
"certifications": ["CE", "FCC", "RoHS", ...],
|
|
"mtbf_hours": number or null,
|
|
"description": "string",
|
|
"notes": ["important notes from the datasheet"],
|
|
"extraction_confidence": 1-10
|
|
}
|
|
|
|
Extraction rules:
|
|
- Tables with Min/Typ/Max columns: extract all three values separately.
|
|
- Docling may create markdown tables with | separators — parse these carefully.
|
|
- Do NOT confuse TX power with RX sensitivity.
|
|
- For multi-channel transceivers (LR4, CWDM4, etc.): list all channels in wavelength_channels.
|
|
- If a field has multiple values (per channel), use the representative or average value and note in notes[].
|
|
- Absolute maximum ratings are NOT the same as operating ranges — use operating specs.
|
|
- Always extract MTBF if listed (often in reliability section at the end).
|
|
- Set extraction_confidence: 10 = full datasheet with all tables intact, 7 = partial extraction, 4 = only partial specs found.
|
|
|
|
{{few_shot_examples}}
|
|
|
|
user_template: |
|
|
Extract all specifications from this datasheet:
|
|
|
|
Part identifier (if known): {{part_hint}}
|
|
Vendor (if known): {{vendor_hint}}
|
|
|
|
Docling-converted Markdown:
|
|
{{datasheet_markdown}}
|
|
|
|
Return JSON only:
|
|
|
|
few_shot_examples:
|
|
- user: |
|
|
Part identifier (if known): QSFP-100G-LR4
|
|
Vendor (if known): Cisco Systems
|
|
|
|
Docling-converted Markdown:
|
|
# QSFP-100G-LR4 Product Specification
|
|
|
|
## General Description
|
|
The QSFP-100G-LR4 is a 4-lane optical transceiver module for 100 Gigabit Ethernet applications.
|
|
It uses four CWDM lanes at 1295.56, 1300.05, 1304.58, and 1309.14 nm.
|
|
|
|
## Optical Specifications
|
|
|
|
| Parameter | Min | Typ | Max | Unit |
|
|
|-----------|-----|-----|-----|------|
|
|
| TX Power per Lane | -6.5 | -3.0 | 2.5 | dBm |
|
|
| RX Sensitivity | — | — | -12.6 | dBm |
|
|
| Center Wavelength Lane 1 | 1294.53 | 1295.56 | 1296.59 | nm |
|
|
| Extinction Ratio | 4.0 | — | — | dB |
|
|
|
|
## Electrical / Environmental
|
|
|
|
| Parameter | Value |
|
|
|-----------|-------|
|
|
| Supply Voltage | 3.3V |
|
|
| Max Power Consumption | 4.0W |
|
|
| Operating Temperature | 0°C to 70°C |
|
|
| Storage Temperature | -40°C to 85°C |
|
|
| MTBF | 1,000,000 hours |
|
|
|
|
## Standards
|
|
- IEEE 802.3ba 100GBASE-LR4
|
|
- SFF-8665 (QSFP28 MSA)
|
|
- SFF-8472 DOM
|
|
|
|
## Certifications: CE, FCC, RoHS
|
|
assistant: |
|
|
{
|
|
"part_number": "QSFP-100G-LR4",
|
|
"vendor": "Cisco Systems",
|
|
"form_factor": "QSFP28",
|
|
"data_rate_gbps": 100,
|
|
"protocol": "Ethernet",
|
|
"wavelength_nm": 1300,
|
|
"wavelength_channels": [1295.56, 1300.05, 1304.58, 1309.14],
|
|
"reach_m": 10000,
|
|
"connector": "LC",
|
|
"fiber_type": "SMF",
|
|
"electrical": {
|
|
"tx_power_min_dbm": -6.5,
|
|
"tx_power_typ_dbm": -3.0,
|
|
"tx_power_max_dbm": 2.5,
|
|
"rx_sensitivity_min_dbm": null,
|
|
"rx_sensitivity_typ_dbm": null,
|
|
"rx_sensitivity_max_dbm": -12.6,
|
|
"extinction_ratio_min_db": 4.0,
|
|
"oma_sensitivity_dbm": null,
|
|
"center_wavelength_min_nm": 1294.53,
|
|
"center_wavelength_max_nm": 1296.59
|
|
},
|
|
"supply_voltage_v": 3.3,
|
|
"supply_current_ma_max": null,
|
|
"power_consumption_w_max": 4.0,
|
|
"temperature_case_min_c": null,
|
|
"temperature_case_max_c": null,
|
|
"temperature_operating_min_c": 0,
|
|
"temperature_operating_max_c": 70,
|
|
"storage_temp_min_c": -40,
|
|
"storage_temp_max_c": 85,
|
|
"humidity_operating_pct_max": null,
|
|
"dom_support": true,
|
|
"standards_compliance": ["IEEE 802.3ba 100GBASE-LR4", "SFF-8665", "SFF-8472"],
|
|
"certifications": ["CE", "FCC", "RoHS"],
|
|
"mtbf_hours": 1000000,
|
|
"description": "4-lane QSFP28 100GBASE-LR4 transceiver using CWDM lanes at 1295-1309 nm for 100 Gbps Ethernet over single-mode fiber up to 10 km.",
|
|
"notes": ["TX power spec is per-lane value", "RX sensitivity is aggregate 100G value"],
|
|
"extraction_confidence": 9
|
|
}
|
|
|
|
variables:
|
|
- part_hint
|
|
- vendor_hint
|
|
- datasheet_markdown
|
|
- few_shot_examples
|
|
|
|
validation_rules:
|
|
tip_validator: true
|
|
min_confidence: 8.0
|
|
required_fields: ["part_number", "form_factor", "data_rate_gbps", "electrical"]
|
|
output_must_be_json: true
|