Librechat - Revision history

Charliep: /* Getting Help */

2026-08-10T11:22:05Z

Getting Help

← Older revision		Revision as of 11:22, 10 August 2026
Line 195:		Line 195:

	= Getting Help =		= Getting Help =
	* Submit a ticket through ~~whateveritiscalled and choose thecorrectoptions and it will be routed to~~ the Librechat ~~support group~~.		* Submit a ticket through the Help Desk's ticket interface for Information Technology -> AI -> Librechat, [https://helpdesk.earlham.edu/TDClient/57/Portal/Requests/ServiceCatalog/Category/837/Artificial-Intelligence found here]

Charliep: /* Building an Agent */

2026-08-10T09:23:29Z

Building an Agent

← Older revision		Revision as of 09:23, 10 August 2026
Line 188:		Line 188:
	You are an Archaeological Prediction Engine specializing in Norse settlement on the island of Newfoundland, Canada. Your job is to analyze text, geographic data, and environmental indicators, and try to guess new locations that are likely to be undiscovered Norse settlements on/around the island of Newfoundland. Evaluate using historical evidence, research documents, and geographic features. A Norse settlement required a protected harbor or bay, access to navigable ocean within 1km, a freshwater source within 1km, bog iron deposits within 3km for iron smelting, forest within 2km for timber, and food access via a cod fishery or salmon river. Using the provided materials, give me a top-50 list of precise coordinates on/around the island of Newfoundland, formatted at the beginning of your response as a Python dictionary, with the name of the location as the key and the coordinates in a list (format [latitude, longitude]) as the value. Latitude and longitude should be provided to the thousandth of a degree (i.e. -20.254, 10.555). Afterward, provide the historical reasoning and geographic viability for each. Every candidate must be on land; if any are in the ocean, they must be rejected immediately and replaced. Known landmarks are L'Anse aux Meadows (confirmed Norse) at 51.596, -55.533, St. John's at 47.561, -52.712, and Cape Race at 46.659, -53.074. If two candidates are within 20km of each other, discard the weaker one and find an alternative. Keep in mind that a degree of latitude is equal to 110 kilometers, and a degree of longitude in this area is equal to 45 kilometers. The Norse would have been traveling either on foot, very slowly, or more quickly if navigable water, like a river, was nearby. As such, locations that are within 1 degree of each other should not be considered "near" each other unless connected by a river or other body of navigable water.		You are an Archaeological Prediction Engine specializing in Norse settlement on the island of Newfoundland, Canada. Your job is to analyze text, geographic data, and environmental indicators, and try to guess new locations that are likely to be undiscovered Norse settlements on/around the island of Newfoundland. Evaluate using historical evidence, research documents, and geographic features. A Norse settlement required a protected harbor or bay, access to navigable ocean within 1km, a freshwater source within 1km, bog iron deposits within 3km for iron smelting, forest within 2km for timber, and food access via a cod fishery or salmon river. Using the provided materials, give me a top-50 list of precise coordinates on/around the island of Newfoundland, formatted at the beginning of your response as a Python dictionary, with the name of the location as the key and the coordinates in a list (format [latitude, longitude]) as the value. Latitude and longitude should be provided to the thousandth of a degree (i.e. -20.254, 10.555). Afterward, provide the historical reasoning and geographic viability for each. Every candidate must be on land; if any are in the ocean, they must be rejected immediately and replaced. Known landmarks are L'Anse aux Meadows (confirmed Norse) at 51.596, -55.533, St. John's at 47.561, -52.712, and Cape Race at 46.659, -53.074. If two candidates are within 20km of each other, discard the weaker one and find an alternative. Keep in mind that a degree of latitude is equal to 110 kilometers, and a degree of longitude in this area is equal to 45 kilometers. The Norse would have been traveling either on foot, very slowly, or more quickly if navigable water, like a river, was nearby. As such, locations that are within 1 degree of each other should not be considered "near" each other unless connected by a river or other body of navigable water.
	</blockquote>		</blockquote>
	<li value="5"> If you would like to add local documents use the Tools menu to add File Search. In the left hand menu you can use the Attach Files icon (paper clip) to upload files to the agent. </li>		# <li value="6">If you would like to add local documents use the Tools menu to add File Search. In the left hand menu you can use the Attach Files icon (paper clip) to upload files to the agent.</li>
	# Add contact information, press Save.		# Add contact information, press Save.
	# Query, evaluate, adjust; repeat. Keep track of the options, documents, queries, and parameter settings you test. Given the number of choices it is easy loose track of what you have done.		# Query, evaluate, adjust; repeat. Keep track of the options, documents, queries, and parameter settings you test. Given the number of choices it is easy loose track of what you have done.

Charliep: /* Building an Agent */

2026-08-10T09:19:44Z

Building an Agent

Charliep: /* Building an Agent */

2026-08-10T09:17:06Z

Building an Agent

Charliep: /* Building an Agent */

2026-08-10T09:16:29Z

Building an Agent

← Older revision		Revision as of 09:16, 10 August 2026
Line 181:		Line 181:

	1. Open Librechat. In the left sidebar find the icon that looks like the overhead view of a toaster "Agent Builder", press that. The left side bar will change to the Create Agent menu.		1. Open Librechat. In the left sidebar find the icon that looks like the overhead view of a toaster "Agent Builder", press that. The left side bar will change to the Create Agent menu.
	2. Give your agent a name and a description.		<li>2. Give your agent a name and a description.
	3. Press the Model choice, the provider should be Ollama (local models). See [[#Local_Models_(Ollama)\|Local Models]] for more information about specific model's capabilities. We choose qwen3-coder-next for its ability to work with spacial data. All the other parameters can be left as-is to begin with.		<li>3. Press the Model choice, the provider should be Ollama (local models). See [[#Local_Models_(Ollama)\|Local Models]] for more information about specific model's capabilities. We choose qwen3-coder-next for its ability to work with spacial data. All the other parameters can be left as-is to begin with.
	4. Use the return arrow (to the left of Model Parameters) to go back to the Create Agent menu. You can choose a category other than General if one of the built-in categories fits your needs, if there is not an obvious fit General is a good choice to begin with.		4. Use the return arrow (to the left of Model Parameters) to go back to the Create Agent menu. You can choose a category other than General if one of the built-in categories fits your needs, if there is not an obvious fit General is a good choice to begin with.
	5. Enter instructions for your agent. These are the guiding principles if you will for the agent's work. Here is an example from the Norse project:		5. Enter instructions for your agent. These are the guiding principles if you will for the agent's work. Here is an example from the Norse project:

Charliep: /* Building an Agent */

2026-08-10T09:16:01Z

Building an Agent

← Older revision		Revision as of 09:16, 10 August 2026
Line 180:		Line 180:
	These are very basic instructions for building an agent using a local model and specific resources. The example used comes from a project looking for other potential Norse inhabitations in Newfoundland in addition to L'Anse aux Meadows.		These are very basic instructions for building an agent using a local model and specific resources. The example used comes from a project looking for other potential Norse inhabitations in Newfoundland in addition to L'Anse aux Meadows.

	# Open Librechat. In the left sidebar find the icon that looks like the overhead view of a toaster "Agent Builder", press that. The left side bar will change to the Create Agent menu.		1. Open Librechat. In the left sidebar find the icon that looks like the overhead view of a toaster "Agent Builder", press that. The left side bar will change to the Create Agent menu.
	# Give your agent a name and a description.		2. Give your agent a name and a description.
	# Press the Model choice, the provider should be Ollama (local models). See [[#Local_Models_(Ollama)\|Local Models]] for more information about specific model's capabilities. We choose qwen3-coder-next for its ability to work with spacial data. All the other parameters can be left as-is to begin with.		3. Press the Model choice, the provider should be Ollama (local models). See [[#Local_Models_(Ollama)\|Local Models]] for more information about specific model's capabilities. We choose qwen3-coder-next for its ability to work with spacial data. All the other parameters can be left as-is to begin with.
	# Use the return arrow (to the left of Model Parameters) to go back to the Create Agent menu. You can choose a category other than General if one of the built-in categories fits your needs, if there is not an obvious fit General is a good choice to begin with.		4. Use the return arrow (to the left of Model Parameters) to go back to the Create Agent menu. You can choose a category other than General if one of the built-in categories fits your needs, if there is not an obvious fit General is a good choice to begin with.
	# Enter instructions for your agent. These are the guiding principles if you will for the agent's work. Here is an example from the Norse project:		5. Enter instructions for your agent. These are the guiding principles if you will for the agent's work. Here is an example from the Norse project:
	<blockquote>		<blockquote>
	You are an Archaeological Prediction Engine specializing in Norse settlement on the island of Newfoundland, Canada. Your job is to analyze text, geographic data, and environmental indicators, and try to guess new locations that are likely to be undiscovered Norse settlements on/around the island of Newfoundland. Evaluate using historical evidence, research documents, and geographic features. A Norse settlement required a protected harbor or bay, access to navigable ocean within 1km, a freshwater source within 1km, bog iron deposits within 3km for iron smelting, forest within 2km for timber, and food access via a cod fishery or salmon river. Using the provided materials, give me a top-50 list of precise coordinates on/around the island of Newfoundland, formatted at the beginning of your response as a Python dictionary, with the name of the location as the key and the coordinates in a list (format [latitude, longitude]) as the value. Latitude and longitude should be provided to the thousandth of a degree (i.e. -20.254, 10.555). Afterward, provide the historical reasoning and geographic viability for each. Every candidate must be on land; if any are in the ocean, they must be rejected immediately and replaced. Known landmarks are L'Anse aux Meadows (confirmed Norse) at 51.596, -55.533, St. John's at 47.561, -52.712, and Cape Race at 46.659, -53.074. If two candidates are within 20km of each other, discard the weaker one and find an alternative. Keep in mind that a degree of latitude is equal to 110 kilometers, and a degree of longitude in this area is equal to 45 kilometers. The Norse would have been traveling either on foot, very slowly, or more quickly if navigable water, like a river, was nearby. As such, locations that are within 1 degree of each other should not be considered "near" each other unless connected by a river or other body of navigable water.		You are an Archaeological Prediction Engine specializing in Norse settlement on the island of Newfoundland, Canada. Your job is to analyze text, geographic data, and environmental indicators, and try to guess new locations that are likely to be undiscovered Norse settlements on/around the island of Newfoundland. Evaluate using historical evidence, research documents, and geographic features. A Norse settlement required a protected harbor or bay, access to navigable ocean within 1km, a freshwater source within 1km, bog iron deposits within 3km for iron smelting, forest within 2km for timber, and food access via a cod fishery or salmon river. Using the provided materials, give me a top-50 list of precise coordinates on/around the island of Newfoundland, formatted at the beginning of your response as a Python dictionary, with the name of the location as the key and the coordinates in a list (format [latitude, longitude]) as the value. Latitude and longitude should be provided to the thousandth of a degree (i.e. -20.254, 10.555). Afterward, provide the historical reasoning and geographic viability for each. Every candidate must be on land; if any are in the ocean, they must be rejected immediately and replaced. Known landmarks are L'Anse aux Meadows (confirmed Norse) at 51.596, -55.533, St. John's at 47.561, -52.712, and Cape Race at 46.659, -53.074. If two candidates are within 20km of each other, discard the weaker one and find an alternative. Keep in mind that a degree of latitude is equal to 110 kilometers, and a degree of longitude in this area is equal to 45 kilometers. The Norse would have been traveling either on foot, very slowly, or more quickly if navigable water, like a river, was nearby. As such, locations that are within 1 degree of each other should not be considered "near" each other unless connected by a river or other body of navigable water.
	</blockquote>		</blockquote>
	~~<li value="X">~~		6. If you would like to add local documents use the Tools menu to add File Search. In the left hand menu you can use the Attach Files icon (paper clip) to upload files to the agent.
	# If you would like to add local documents use the Tools menu to add File Search. In the left hand menu you can use the Attach Files icon (paper clip) to upload files to the agent.		7. Add contact information, press Save.
	~~# Press~~ Save.		8. Query, evaluate, adjust; repeat. Keep track of the options, documents, queries, and parameter settings you test. Given the number of choices it is easy loose track of what you have done.
	# Query, evaluate, adjust; repeat. Keep track of the options, documents, queries, and parameter settings you test. Given the number of choices it is easy loose track of what you have done.

	There are many ways to tune an agent based on your testing and usage. Two places to start are with the category and the temperature parameter for a model.		There are many ways to tune an agent based on your testing and usage. Two places to start are with the category and the temperature parameter for a model.

Charliep: /* Building an Agent */

2026-08-10T09:14:23Z

Building an Agent

← Older revision		Revision as of 09:14, 10 August 2026
Line 188:		Line 188:
	You are an Archaeological Prediction Engine specializing in Norse settlement on the island of Newfoundland, Canada. Your job is to analyze text, geographic data, and environmental indicators, and try to guess new locations that are likely to be undiscovered Norse settlements on/around the island of Newfoundland. Evaluate using historical evidence, research documents, and geographic features. A Norse settlement required a protected harbor or bay, access to navigable ocean within 1km, a freshwater source within 1km, bog iron deposits within 3km for iron smelting, forest within 2km for timber, and food access via a cod fishery or salmon river. Using the provided materials, give me a top-50 list of precise coordinates on/around the island of Newfoundland, formatted at the beginning of your response as a Python dictionary, with the name of the location as the key and the coordinates in a list (format [latitude, longitude]) as the value. Latitude and longitude should be provided to the thousandth of a degree (i.e. -20.254, 10.555). Afterward, provide the historical reasoning and geographic viability for each. Every candidate must be on land; if any are in the ocean, they must be rejected immediately and replaced. Known landmarks are L'Anse aux Meadows (confirmed Norse) at 51.596, -55.533, St. John's at 47.561, -52.712, and Cape Race at 46.659, -53.074. If two candidates are within 20km of each other, discard the weaker one and find an alternative. Keep in mind that a degree of latitude is equal to 110 kilometers, and a degree of longitude in this area is equal to 45 kilometers. The Norse would have been traveling either on foot, very slowly, or more quickly if navigable water, like a river, was nearby. As such, locations that are within 1 degree of each other should not be considered "near" each other unless connected by a river or other body of navigable water.		You are an Archaeological Prediction Engine specializing in Norse settlement on the island of Newfoundland, Canada. Your job is to analyze text, geographic data, and environmental indicators, and try to guess new locations that are likely to be undiscovered Norse settlements on/around the island of Newfoundland. Evaluate using historical evidence, research documents, and geographic features. A Norse settlement required a protected harbor or bay, access to navigable ocean within 1km, a freshwater source within 1km, bog iron deposits within 3km for iron smelting, forest within 2km for timber, and food access via a cod fishery or salmon river. Using the provided materials, give me a top-50 list of precise coordinates on/around the island of Newfoundland, formatted at the beginning of your response as a Python dictionary, with the name of the location as the key and the coordinates in a list (format [latitude, longitude]) as the value. Latitude and longitude should be provided to the thousandth of a degree (i.e. -20.254, 10.555). Afterward, provide the historical reasoning and geographic viability for each. Every candidate must be on land; if any are in the ocean, they must be rejected immediately and replaced. Known landmarks are L'Anse aux Meadows (confirmed Norse) at 51.596, -55.533, St. John's at 47.561, -52.712, and Cape Race at 46.659, -53.074. If two candidates are within 20km of each other, discard the weaker one and find an alternative. Keep in mind that a degree of latitude is equal to 110 kilometers, and a degree of longitude in this area is equal to 45 kilometers. The Norse would have been traveling either on foot, very slowly, or more quickly if navigable water, like a river, was nearby. As such, locations that are within 1 degree of each other should not be considered "near" each other unless connected by a river or other body of navigable water.
	</blockquote>		</blockquote>
	<li value="5">		<li value="X">
	# If you would like to add local documents use the Tools menu to add File Search. In the left hand menu you can use the Attach Files icon (paper clip) to upload files to the agent.		# If you would like to add local documents use the Tools menu to add File Search. In the left hand menu you can use the Attach Files icon (paper clip) to upload files to the agent.
	# Press Save.		# Press Save.

Charliep: /* Building an Agent */

2026-08-10T09:14:06Z

Building an Agent

← Older revision		Revision as of 09:14, 10 August 2026
Line 188:		Line 188:
	You are an Archaeological Prediction Engine specializing in Norse settlement on the island of Newfoundland, Canada. Your job is to analyze text, geographic data, and environmental indicators, and try to guess new locations that are likely to be undiscovered Norse settlements on/around the island of Newfoundland. Evaluate using historical evidence, research documents, and geographic features. A Norse settlement required a protected harbor or bay, access to navigable ocean within 1km, a freshwater source within 1km, bog iron deposits within 3km for iron smelting, forest within 2km for timber, and food access via a cod fishery or salmon river. Using the provided materials, give me a top-50 list of precise coordinates on/around the island of Newfoundland, formatted at the beginning of your response as a Python dictionary, with the name of the location as the key and the coordinates in a list (format [latitude, longitude]) as the value. Latitude and longitude should be provided to the thousandth of a degree (i.e. -20.254, 10.555). Afterward, provide the historical reasoning and geographic viability for each. Every candidate must be on land; if any are in the ocean, they must be rejected immediately and replaced. Known landmarks are L'Anse aux Meadows (confirmed Norse) at 51.596, -55.533, St. John's at 47.561, -52.712, and Cape Race at 46.659, -53.074. If two candidates are within 20km of each other, discard the weaker one and find an alternative. Keep in mind that a degree of latitude is equal to 110 kilometers, and a degree of longitude in this area is equal to 45 kilometers. The Norse would have been traveling either on foot, very slowly, or more quickly if navigable water, like a river, was nearby. As such, locations that are within 1 degree of each other should not be considered "near" each other unless connected by a river or other body of navigable water.		You are an Archaeological Prediction Engine specializing in Norse settlement on the island of Newfoundland, Canada. Your job is to analyze text, geographic data, and environmental indicators, and try to guess new locations that are likely to be undiscovered Norse settlements on/around the island of Newfoundland. Evaluate using historical evidence, research documents, and geographic features. A Norse settlement required a protected harbor or bay, access to navigable ocean within 1km, a freshwater source within 1km, bog iron deposits within 3km for iron smelting, forest within 2km for timber, and food access via a cod fishery or salmon river. Using the provided materials, give me a top-50 list of precise coordinates on/around the island of Newfoundland, formatted at the beginning of your response as a Python dictionary, with the name of the location as the key and the coordinates in a list (format [latitude, longitude]) as the value. Latitude and longitude should be provided to the thousandth of a degree (i.e. -20.254, 10.555). Afterward, provide the historical reasoning and geographic viability for each. Every candidate must be on land; if any are in the ocean, they must be rejected immediately and replaced. Known landmarks are L'Anse aux Meadows (confirmed Norse) at 51.596, -55.533, St. John's at 47.561, -52.712, and Cape Race at 46.659, -53.074. If two candidates are within 20km of each other, discard the weaker one and find an alternative. Keep in mind that a degree of latitude is equal to 110 kilometers, and a degree of longitude in this area is equal to 45 kilometers. The Norse would have been traveling either on foot, very slowly, or more quickly if navigable water, like a river, was nearby. As such, locations that are within 1 degree of each other should not be considered "near" each other unless connected by a river or other body of navigable water.
	</blockquote>		</blockquote>
			<li value="5">
	# If you would like to add local documents use the Tools menu to add File Search. In the left hand menu you can use the Attach Files icon (paper clip) to upload files to the agent.		# If you would like to add local documents use the Tools menu to add File Search. In the left hand menu you can use the Attach Files icon (paper clip) to upload files to the agent.
	# Press Save.		# Press Save.

Charliep: /* Building an Agent */

2026-08-10T09:12:45Z

Building an Agent

← Older revision		Revision as of 09:12, 10 August 2026
Line 183:		Line 183:
	# Give your agent a name and a description.		# Give your agent a name and a description.
	# Press the Model choice, the provider should be Ollama (local models). See [[#Local_Models_(Ollama)\|Local Models]] for more information about specific model's capabilities. We choose qwen3-coder-next for its ability to work with spacial data. All the other parameters can be left as-is to begin with.		# Press the Model choice, the provider should be Ollama (local models). See [[#Local_Models_(Ollama)\|Local Models]] for more information about specific model's capabilities. We choose qwen3-coder-next for its ability to work with spacial data. All the other parameters can be left as-is to begin with.
	# Use the return arrow (to the left of Model Parameters) to go back to the Create Agent menu. You can choose a category other than General if one fits your needs, to begin with ~~General~~ is a ~~good choice~~.		# Use the return arrow (to the left of Model Parameters) to go back to the Create Agent menu. You can choose a category other than General if one of the built-in categories fits your needs, if there is not an obvious fit General is a good choice to begin with.
			# Enter instructions for your agent. These are the guiding principles if you will for the agent's work. Here is an example from the Norse project:
			<blockquote>
			You are an Archaeological Prediction Engine specializing in Norse settlement on the island of Newfoundland, Canada. Your job is to analyze text, geographic data, and environmental indicators, and try to guess new locations that are likely to be undiscovered Norse settlements on/around the island of Newfoundland. Evaluate using historical evidence, research documents, and geographic features. A Norse settlement required a protected harbor or bay, access to navigable ocean within 1km, a freshwater source within 1km, bog iron deposits within 3km for iron smelting, forest within 2km for timber, and food access via a cod fishery or salmon river. Using the provided materials, give me a top-50 list of precise coordinates on/around the island of Newfoundland, formatted at the beginning of your response as a Python dictionary, with the name of the location as the key and the coordinates in a list (format [latitude, longitude]) as the value. Latitude and longitude should be provided to the thousandth of a degree (i.e. -20.254, 10.555). Afterward, provide the historical reasoning and geographic viability for each. Every candidate must be on land; if any are in the ocean, they must be rejected immediately and replaced. Known landmarks are L'Anse aux Meadows (confirmed Norse) at 51.596, -55.533, St. John's at 47.561, -52.712, and Cape Race at 46.659, -53.074. If two candidates are within 20km of each other, discard the weaker one and find an alternative. Keep in mind that a degree of latitude is equal to 110 kilometers, and a degree of longitude in this area is equal to 45 kilometers. The Norse would have been traveling either on foot, very slowly, or more quickly if navigable water, like a river, was nearby. As such, locations that are within 1 degree of each other should not be considered "near" each other unless connected by a river or other body of navigable water.
			</blockquote>
			# If you would like to add local documents use the Tools menu to add File Search. In the left hand menu you can use the Attach Files icon (paper clip) to upload files to the agent.
			# Press Save.
			# Query, evaluate, adjust; repeat. Keep track of the options, documents, queries, and parameter settings you test. Given the number of choices it is easy loose track of what you have done.

	There are many ways to tune an agent based on your testing and usage. Two places to start are with the category and the temperature parameter for a model.		There are many ways to tune an agent based on your testing and usage. Two places to start are with the category and the temperature parameter for a model.

Charliep: /* Building an agent */

2026-08-10T01:05:30Z

Building an agent

← Older revision		Revision as of 01:05, 10 August 2026
Line 177:		Line 177:
	* Save prompts you like. If you craft a really effective prompt, save it so you can reuse it easily.		* Save prompts you like. If you craft a really effective prompt, save it so you can reuse it easily.

	= Building an ~~agent~~ =		= Building an Agent =
	These are very basic instructions for building an agent using a local model and specific resources.		These are very basic instructions for building an agent using a local model and specific resources. The example used comes from a project looking for other potential Norse inhabitations in Newfoundland in addition to L'Anse aux Meadows.
	* Open Librechat. In the left sidebar find the icon that looks like the overhead view of a toaster "Agent Builder", press that.
	* more		# Open Librechat. In the left sidebar find the icon that looks like the overhead view of a toaster "Agent Builder", press that. The left side bar will change to the Create Agent menu.
			# Give your agent a name and a description.
			# Press the Model choice, the provider should be Ollama (local models). See [[#Local_Models_(Ollama)\|Local Models]] for more information about specific model's capabilities. We choose qwen3-coder-next for its ability to work with spacial data. All the other parameters can be left as-is to begin with.
			# Use the return arrow (to the left of Model Parameters) to go back to the Create Agent menu. You can choose a category other than General if one fits your needs, to begin with General is a good choice.

			There are many ways to tune an agent based on your testing and usage. Two places to start are with the category and the temperature parameter for a model.

	= Getting Help =		= Getting Help =
	* Submit a ticket through whateveritiscalled and choose thecorrectoptions and it will be routed to the Librechat support group.		* Submit a ticket through whateveritiscalled and choose thecorrectoptions and it will be routed to the Librechat support group.

← Older revision		Revision as of 09:19, 10 August 2026
Line 180:		Line 180:
	These are very basic instructions for building an agent using a local model and specific resources. The example used comes from a project looking for other potential Norse inhabitations in Newfoundland in addition to L'Anse aux Meadows.		These are very basic instructions for building an agent using a local model and specific resources. The example used comes from a project looking for other potential Norse inhabitations in Newfoundland in addition to L'Anse aux Meadows.

	~~<li>1.~~ Open Librechat. In the left sidebar find the icon that looks like the overhead view of a toaster "Agent Builder", press that. The left side bar will change to the Create Agent menu.		# Open Librechat. In the left sidebar find the icon that looks like the overhead view of a toaster "Agent Builder", press that. The left side bar will change to the Create Agent menu.
	~~<li>2.~~ Give your agent a name and a description.		# Give your agent a name and a description.
	~~<li>3.~~ Press the Model choice, the provider should be Ollama (local models). See [[#Local_Models_(Ollama)\|Local Models]] for more information about specific model's capabilities. We choose qwen3-coder-next for its ability to work with spacial data. All the other parameters can be left as-is to begin with.		# Press the Model choice, the provider should be Ollama (local models). See [[#Local_Models_(Ollama)\|Local Models]] for more information about specific model's capabilities. We choose qwen3-coder-next for its ability to work with spacial data. All the other parameters can be left as-is to begin with.
	~~<li>4.~~ Use the return arrow (to the left of Model Parameters) to go back to the Create Agent menu. You can choose a category other than General if one of the built-in categories fits your needs, if there is not an obvious fit General is a good choice to begin with.		# Use the return arrow (to the left of Model Parameters) to go back to the Create Agent menu. You can choose a category other than General if one of the built-in categories fits your needs, if there is not an obvious fit General is a good choice to begin with.
	~~<li>5.~~ Enter instructions for your agent. These are the guiding principles if you will for the agent's work. Here is an example from the Norse project:		# Enter instructions for your agent. These are the guiding principles if you will for the agent's work. Here is an example from the Norse project:
	<blockquote>		<blockquote>
	You are an Archaeological Prediction Engine specializing in Norse settlement on the island of Newfoundland, Canada. Your job is to analyze text, geographic data, and environmental indicators, and try to guess new locations that are likely to be undiscovered Norse settlements on/around the island of Newfoundland. Evaluate using historical evidence, research documents, and geographic features. A Norse settlement required a protected harbor or bay, access to navigable ocean within 1km, a freshwater source within 1km, bog iron deposits within 3km for iron smelting, forest within 2km for timber, and food access via a cod fishery or salmon river. Using the provided materials, give me a top-50 list of precise coordinates on/around the island of Newfoundland, formatted at the beginning of your response as a Python dictionary, with the name of the location as the key and the coordinates in a list (format [latitude, longitude]) as the value. Latitude and longitude should be provided to the thousandth of a degree (i.e. -20.254, 10.555). Afterward, provide the historical reasoning and geographic viability for each. Every candidate must be on land; if any are in the ocean, they must be rejected immediately and replaced. Known landmarks are L'Anse aux Meadows (confirmed Norse) at 51.596, -55.533, St. John's at 47.561, -52.712, and Cape Race at 46.659, -53.074. If two candidates are within 20km of each other, discard the weaker one and find an alternative. Keep in mind that a degree of latitude is equal to 110 kilometers, and a degree of longitude in this area is equal to 45 kilometers. The Norse would have been traveling either on foot, very slowly, or more quickly if navigable water, like a river, was nearby. As such, locations that are within 1 degree of each other should not be considered "near" each other unless connected by a river or other body of navigable water.		You are an Archaeological Prediction Engine specializing in Norse settlement on the island of Newfoundland, Canada. Your job is to analyze text, geographic data, and environmental indicators, and try to guess new locations that are likely to be undiscovered Norse settlements on/around the island of Newfoundland. Evaluate using historical evidence, research documents, and geographic features. A Norse settlement required a protected harbor or bay, access to navigable ocean within 1km, a freshwater source within 1km, bog iron deposits within 3km for iron smelting, forest within 2km for timber, and food access via a cod fishery or salmon river. Using the provided materials, give me a top-50 list of precise coordinates on/around the island of Newfoundland, formatted at the beginning of your response as a Python dictionary, with the name of the location as the key and the coordinates in a list (format [latitude, longitude]) as the value. Latitude and longitude should be provided to the thousandth of a degree (i.e. -20.254, 10.555). Afterward, provide the historical reasoning and geographic viability for each. Every candidate must be on land; if any are in the ocean, they must be rejected immediately and replaced. Known landmarks are L'Anse aux Meadows (confirmed Norse) at 51.596, -55.533, St. John's at 47.561, -52.712, and Cape Race at 46.659, -53.074. If two candidates are within 20km of each other, discard the weaker one and find an alternative. Keep in mind that a degree of latitude is equal to 110 kilometers, and a degree of longitude in this area is equal to 45 kilometers. The Norse would have been traveling either on foot, very slowly, or more quickly if navigable water, like a river, was nearby. As such, locations that are within 1 degree of each other should not be considered "near" each other unless connected by a river or other body of navigable water.
	</blockquote>		</blockquote>
	<li>6. If you would like to add local documents use the Tools menu to add File Search. In the left hand menu you can use the Attach Files icon (paper clip) to upload files to the agent.		<li value="5"> If you would like to add local documents use the Tools menu to add File Search. In the left hand menu you can use the Attach Files icon (paper clip) to upload files to the agent. </li>
	<li>7. Add contact information, press Save.		# Add contact information, press Save.
	~~<li>8.~~ Query, evaluate, adjust; repeat. Keep track of the options, documents, queries, and parameter settings you test. Given the number of choices it is easy loose track of what you have done.		# Query, evaluate, adjust; repeat. Keep track of the options, documents, queries, and parameter settings you test. Given the number of choices it is easy loose track of what you have done.

	There are many ways to tune an agent based on your testing and usage. Two places to start are with the category and the temperature parameter for a model.		There are many ways to tune an agent based on your testing and usage. Two places to start are with the category and the temperature parameter for a model.

← Older revision		Revision as of 09:17, 10 August 2026
Line 180:		Line 180:
	These are very basic instructions for building an agent using a local model and specific resources. The example used comes from a project looking for other potential Norse inhabitations in Newfoundland in addition to L'Anse aux Meadows.		These are very basic instructions for building an agent using a local model and specific resources. The example used comes from a project looking for other potential Norse inhabitations in Newfoundland in addition to L'Anse aux Meadows.

	1. Open Librechat. In the left sidebar find the icon that looks like the overhead view of a toaster "Agent Builder", press that. The left side bar will change to the Create Agent menu.		<li>1. Open Librechat. In the left sidebar find the icon that looks like the overhead view of a toaster "Agent Builder", press that. The left side bar will change to the Create Agent menu.
	<li>2. Give your agent a name and a description.		<li>2. Give your agent a name and a description.
	<li>3. Press the Model choice, the provider should be Ollama (local models). See [[#Local_Models_(Ollama)\|Local Models]] for more information about specific model's capabilities. We choose qwen3-coder-next for its ability to work with spacial data. All the other parameters can be left as-is to begin with.		<li>3. Press the Model choice, the provider should be Ollama (local models). See [[#Local_Models_(Ollama)\|Local Models]] for more information about specific model's capabilities. We choose qwen3-coder-next for its ability to work with spacial data. All the other parameters can be left as-is to begin with.
	4. Use the return arrow (to the left of Model Parameters) to go back to the Create Agent menu. You can choose a category other than General if one of the built-in categories fits your needs, if there is not an obvious fit General is a good choice to begin with.		<li>4. Use the return arrow (to the left of Model Parameters) to go back to the Create Agent menu. You can choose a category other than General if one of the built-in categories fits your needs, if there is not an obvious fit General is a good choice to begin with.
	5. Enter instructions for your agent. These are the guiding principles if you will for the agent's work. Here is an example from the Norse project:		<li>5. Enter instructions for your agent. These are the guiding principles if you will for the agent's work. Here is an example from the Norse project:
	<blockquote>		<blockquote>
	You are an Archaeological Prediction Engine specializing in Norse settlement on the island of Newfoundland, Canada. Your job is to analyze text, geographic data, and environmental indicators, and try to guess new locations that are likely to be undiscovered Norse settlements on/around the island of Newfoundland. Evaluate using historical evidence, research documents, and geographic features. A Norse settlement required a protected harbor or bay, access to navigable ocean within 1km, a freshwater source within 1km, bog iron deposits within 3km for iron smelting, forest within 2km for timber, and food access via a cod fishery or salmon river. Using the provided materials, give me a top-50 list of precise coordinates on/around the island of Newfoundland, formatted at the beginning of your response as a Python dictionary, with the name of the location as the key and the coordinates in a list (format [latitude, longitude]) as the value. Latitude and longitude should be provided to the thousandth of a degree (i.e. -20.254, 10.555). Afterward, provide the historical reasoning and geographic viability for each. Every candidate must be on land; if any are in the ocean, they must be rejected immediately and replaced. Known landmarks are L'Anse aux Meadows (confirmed Norse) at 51.596, -55.533, St. John's at 47.561, -52.712, and Cape Race at 46.659, -53.074. If two candidates are within 20km of each other, discard the weaker one and find an alternative. Keep in mind that a degree of latitude is equal to 110 kilometers, and a degree of longitude in this area is equal to 45 kilometers. The Norse would have been traveling either on foot, very slowly, or more quickly if navigable water, like a river, was nearby. As such, locations that are within 1 degree of each other should not be considered "near" each other unless connected by a river or other body of navigable water.		You are an Archaeological Prediction Engine specializing in Norse settlement on the island of Newfoundland, Canada. Your job is to analyze text, geographic data, and environmental indicators, and try to guess new locations that are likely to be undiscovered Norse settlements on/around the island of Newfoundland. Evaluate using historical evidence, research documents, and geographic features. A Norse settlement required a protected harbor or bay, access to navigable ocean within 1km, a freshwater source within 1km, bog iron deposits within 3km for iron smelting, forest within 2km for timber, and food access via a cod fishery or salmon river. Using the provided materials, give me a top-50 list of precise coordinates on/around the island of Newfoundland, formatted at the beginning of your response as a Python dictionary, with the name of the location as the key and the coordinates in a list (format [latitude, longitude]) as the value. Latitude and longitude should be provided to the thousandth of a degree (i.e. -20.254, 10.555). Afterward, provide the historical reasoning and geographic viability for each. Every candidate must be on land; if any are in the ocean, they must be rejected immediately and replaced. Known landmarks are L'Anse aux Meadows (confirmed Norse) at 51.596, -55.533, St. John's at 47.561, -52.712, and Cape Race at 46.659, -53.074. If two candidates are within 20km of each other, discard the weaker one and find an alternative. Keep in mind that a degree of latitude is equal to 110 kilometers, and a degree of longitude in this area is equal to 45 kilometers. The Norse would have been traveling either on foot, very slowly, or more quickly if navigable water, like a river, was nearby. As such, locations that are within 1 degree of each other should not be considered "near" each other unless connected by a river or other body of navigable water.
	</blockquote>		</blockquote>
	6. If you would like to add local documents use the Tools menu to add File Search. In the left hand menu you can use the Attach Files icon (paper clip) to upload files to the agent.		<li>6. If you would like to add local documents use the Tools menu to add File Search. In the left hand menu you can use the Attach Files icon (paper clip) to upload files to the agent.
	7. Add contact information, press Save.		<li>7. Add contact information, press Save.
	8. Query, evaluate, adjust; repeat. Keep track of the options, documents, queries, and parameter settings you test. Given the number of choices it is easy loose track of what you have done.		<li>8. Query, evaluate, adjust; repeat. Keep track of the options, documents, queries, and parameter settings you test. Given the number of choices it is easy loose track of what you have done.

	There are many ways to tune an agent based on your testing and usage. Two places to start are with the category and the temperature parameter for a model.		There are many ways to tune an agent based on your testing and usage. Two places to start are with the category and the temperature parameter for a model.