"""Solar ordinance decision tree graph setup functions"""
from compass.common import (
setup_graph_no_nodes,
llm_response_starts_with_yes,
llm_response_starts_with_no,
)
[docs]
def setup_graph_sef_types(**kwargs):
"""Setup graph to get the largest solar farm size in the text
Parameters
----------
**kwargs
Keyword-value pairs to add to graph.
Returns
-------
networkx.DiGraph
Graph instance that can be used to initialize an
`elm.tree.DecisionTree`.
"""
G = setup_graph_no_nodes( # noqa: N806
d_tree_name="Solar Energy Farm types", **kwargs
)
G.add_node(
"init",
prompt=(
"Does the following text distinguish between multiple solar "
"energy farm sizes? Distinctions are often made as 'small', "
"'personal', or 'private' vs 'large', 'commercial', or 'utility'. "
"Sometimes the distinction uses actual MW values. "
"Please start your response with either 'Yes' or 'No' and briefly "
"explain your answer."
'\n\n"""\n{text}\n"""'
),
)
G.add_edge("init", "get_text", condition=llm_response_starts_with_yes)
G.add_node(
"get_text",
prompt=(
"What are the different solar energy farm sizes **regulated by "
"this ordinance**? List them in order of increasing size. "
"Include any relevant numerical qualifiers in the name, if "
"appropriate. Only include systems that resemble ground-mounted "
"solar energy farms; do not include other solar energy system "
"types lice CSP or roof-mounted systems or other technologies "
"like wind energy systems, geothermal energy systems, etc."
),
)
G.add_edge("get_text", "get_regulated")
G.add_node(
"get_regulated",
prompt=(
"Are any of these systems **not** regulated by this ordinance?"
),
)
G.add_edge("get_regulated", "get_largest")
G.add_node(
"get_largest",
prompt=(
"What is the **largest** solar energy farm size that **is "
"regulated by this ordinance**?"
),
)
G.add_edge("get_largest", "check_matches_definition")
G.add_node(
"check_matches_definition",
prompt=(
"Does the ordinance explicitly define this system as large, "
"commercial, utility-scale, or something akin to that? "
"Please start your response with either 'Yes' or 'No' and briefly "
"explain your answer."
),
)
G.add_edge(
"check_matches_definition",
"final_large",
condition=llm_response_starts_with_yes,
)
G.add_edge(
"check_matches_definition",
"check_scale_reason",
condition=llm_response_starts_with_no,
)
G.add_node(
"check_scale_reason",
prompt=(
"Would a reasonable person classify this kind of system as a "
"**large**, commercial, or even utility-scale** solar energy farm "
"(e.g. with the primary purpose of generating electricity for "
"sale, as opposed to small, residential, roof-mounted, private, "
"or other kinds of 'small' systems)? "
"Please start your response with either 'Yes' or 'No' and briefly "
"explain your answer."
),
)
G.add_edge(
"check_scale_reason",
"final_large",
condition=llm_response_starts_with_yes,
)
G.add_edge(
"check_scale_reason",
"final_small",
condition=llm_response_starts_with_no,
)
G.add_node(
"final_large",
prompt=(
"Respond based on our entire conversation so far. Return your "
"answer as a dictionary in JSON format (not markdown). Your "
"JSON file must include exactly two keys. The keys are "
"'largest_sef_type' and 'explanation', and 'is_large'. The value "
"of the 'largest_sef_type' key should be a string that labels the "
"largest solar energy system size **regulated by this "
"ordinance**. The value of the 'explanation' key should be a "
"string containing a short explanation for your choice. The value "
"of the 'is_large' key should be the boolean value `true`, since "
"we determined this is a large-scale system."
),
)
G.add_node(
"final_small",
prompt=(
"Respond based on our entire conversation so far. Return your "
"answer as a dictionary in JSON format (not markdown). Your "
"JSON file must include exactly two keys. The keys are "
"'largest_sef_type' and 'explanation', and 'is_large'. The value "
"of the 'largest_sef_type' key should be a string that labels the "
"largest solar energy system size **regulated by this "
"ordinance**. The value of the 'explanation' key should be a "
"string containing a short explanation for your choice. The value "
"of the 'is_large' key should be the boolean value `false`, since "
"we determined this is not a large-scale system."
),
)
return G
[docs]
def setup_multiplier(**kwargs):
"""Setup graph to extract a setbacks multiplier values for a feature
Parameters
----------
**kwargs
Keyword-value pairs to add to graph.
Returns
-------
networkx.DiGraph
Graph instance that can be used to initialize an
`elm.tree.DecisionTree`.
"""
G = setup_graph_no_nodes( # noqa: N806
d_tree_name="Setback distance", **kwargs
)
G.add_node(
"init",
prompt=(
"Does the text mention a multiplier that should be applied to the "
"structure height to compute the setback distance from {feature} "
"for {tech}? "
"Please consider only {feature}; do not respond based on any text "
"related to {ignore_features}. "
"Please also only consider setbacks specifically for "
"{system_size_reminder}"
"Please start your response with either 'Yes' or 'No' and briefly "
"explain your answer."
),
)
G.add_edge("init", "no_multiplier", condition=llm_response_starts_with_no)
G.add_node(
"no_multiplier",
prompt=(
"Does the ordinance give the setback from {feature} as a fixed "
"distance value? "
"Please consider only on {feature}; do not respond based on any "
"text related to {ignore_features}. "
"Please also only consider setbacks specifically for "
"{system_size_reminder}"
"Please start your response with either 'Yes' or "
"'No' and briefly explain your answer."
),
)
G.add_edge(
"no_multiplier", "units", condition=llm_response_starts_with_yes
)
G.add_edge(
"no_multiplier", "out_static", condition=llm_response_starts_with_no
)
G.add_node(
"units",
prompt=(
"What are the units for the setback from {feature}? "
"Ensure that:\n1) You accurately identify the unit value "
"associated with the setback.\n2) The unit is "
"expressed using standard, conventional unit names (e.g., "
"'feet', 'meters', 'miles' etc.)\n3) If multiple "
"values are mentioned, return only the units for the most "
"restrictive value that directly pertains to the setback.\n\n"
"Example Inputs and Outputs:\n"
"Text: 'All WECS Towers shall be set back a distance of at least "
"one thousand (1000) feet, from any primary structure'\n"
"Output: 'feet'\n"
),
)
G.add_edge("units", "out_static")
G.add_node(
"out_static",
prompt=(
"Please respond based on our entire conversation so far. "
"Return your answer in JSON "
"format (not markdown). Your JSON file must include exactly "
"four keys. The keys are 'value', 'units', 'summary', and "
"'section'. The value of the 'value' key should be a "
"**numerical** value corresponding to the setback distance value "
"from {feature} or `null` if there was no such value. The value "
"of the 'units' key should be a string corresponding to the "
"(standard) units of the setback distance value from {feature} "
"or `null` if there was no such value. "
"As before, focus only on setbacks that would apply for"
"{system_size_reminder}"
"{SUMMARY_PROMPT} {SECTION_PROMPT}"
),
)
G.add_edge("init", "m_single", condition=llm_response_starts_with_yes)
G.add_node(
"m_single",
prompt=(
"Are multiple values given for the multiplier used to "
"compute the setback distance value from {feature} for {tech}? "
"Remember to ignore any text related to {ignore_features}. "
"Please only consider setbacks specifically for "
"{system_size_reminder}"
"If so, select and state the largest one. Otherwise, repeat the "
"single multiplier value that was given in the text. "
),
)
G.add_edge("m_single", "adder")
G.add_node(
"adder",
prompt=(
"Does the ordinance for the setback from {feature} include a "
"static distance value that should be added to the result of "
"the multiplication? "
"Remember to ignore any text related to {ignore_features}. "
"Please only consider setbacks specifically for "
"{system_size_reminder}"
"Do not confuse this value with static setback requirements. "
"Ignore text with clauses such as 'no lesser than', 'no greater "
"than', 'the lesser of', or 'the greater of'. Please start your "
"response with either 'Yes' or 'No' and briefly explain your "
"answer, stating the adder value if it exists."
),
)
G.add_edge("adder", "out_no_adder", condition=llm_response_starts_with_no)
G.add_edge("adder", "adder_eq", condition=llm_response_starts_with_yes)
G.add_node(
"adder_eq",
prompt=(
"Does the adder value you identified satisfy the following "
"equation: `multiplier * height + <adder>`? Begin your "
"response with either 'Yes' or 'No' and briefly explain your "
"answer."
),
)
G.add_edge(
"adder_eq", "out_no_adder", condition=llm_response_starts_with_no
)
G.add_edge("adder_eq", "out_m", condition=llm_response_starts_with_no)
G.add_edge(
"adder_eq",
"conversion",
condition=llm_response_starts_with_yes,
)
G.add_node(
"conversion",
prompt=(
"If the adder value is not given in feet, convert "
"it to feet (remember that there are 3.28084 feet in one meter "
"and 5280 feet in one mile). Show your work step-by-step "
"if you had to perform a conversion."
),
)
G.add_edge("conversion", "out_m")
G.add_node(
"out_m",
prompt=(
"Please respond based on our entire conversation so far. "
"Return your answer as a single dictionary in JSON "
"format (not markdown). Your JSON file must include exactly four "
"keys. The keys are 'mult_value', 'adder', 'summary', and "
"'section'. The value of the 'mult_value' key should be a "
"**numerical** value corresponding to the multiplier value we "
"determined earlier. The value of the 'adder' key should be a "
"**numerical** value corresponding to the static value to be "
"added to the total setback distance after multiplication, as we "
"determined earlier, or `null` if there is no such value. "
"{SUMMARY_PROMPT} {SECTION_PROMPT}"
),
)
G.add_node(
"out_no_adder",
prompt=(
"Please respond based on our entire conversation so far. "
"Return your answer as a single dictionary in JSON "
"format (not markdown). Your JSON file must include exactly three "
"keys. The keys are 'mult_value', 'summary', and 'section'. The "
"value of the 'mult_value' key should be a **numerical** value "
"corresponding to the multiplier value we determined earlier. "
"{SUMMARY_PROMPT} {SECTION_PROMPT}"
),
)
return G