Generating custom lighting recipe via LLM by deviating from exemplars

Abstract

A method for generating a lighting recipe is provided. The method includes retrieving an exemplar from an exemplar repository comprising a plurality of exemplars. The exemplar is retrieved based on a user prompt and a current status data set. Each of the plurality of exemplars comprises an optimal recipe, an exemplar prompt, and an exemplar status data set. The method further includes determining one or more deviation parameters based on a prompt deviation between the user prompt and the exemplar prompt of the retrieved exemplar and / or a status deviation between the current status data set and the exemplar status data set of the retrieved exemplar. The method further includes determining one or more allowable recipe adjustment limits based on the one or more deviation parameters. The method further includes generating, via an LLM, the lighting recipe based on the optimal recipe and the one or more allowable recipe adjustment limits.

Description

[0001] 2024PF80360

[0002] 1

[0003] Generating custom lighting recipe via LLM by deviating from exemplars

[0004] FIELD OF THE INVENTION

[0005] The present disclosure is generally directed to generating lighting recipes using large language models (LLMs), and, more particularly, to adjusting exemplars used by LLMs to generate customized lighting recipes.

[0006] BACKGROUND OF THE INVENTION

[0007] Off-the-shelf large language models (LLMs) such as GPT-4 are trained on Internet-scale data. Accordingly, these off-the-shelf LLMs lack a sufficiently deep understanding of the lighting domain to be able to accurately finetune lighting recipes based on user prompts and / or current status data without further guidance.

[0008] SUMMARY OF THE INVENTION

[0009] The present disclosure is generally directed to systems and methods for generating a lighting recipe for one or more luminaires of a lighting system installed in an environment. The system receives a user prompt regarding a desired lighting recipe (such as “Give me happy lighting”) and a current status data set corresponding to the status of the environment, the lighting system configuration, and / or the user. The status of the environment could include information regarding date, time, temperature, ambient lighting, etc. The status of the system configuration could include luminaire count, brightness, color, correlated color temperature, active static and / or dynamic lighting effects, audio outputs, etc. The user prompt may be received from a user interface or other aspect of the system. Further, environmental aspects of the current status data set, such as temperature or occupancy may be captured by one or more sensors arranged in the environment.

[0010] The system retrieves an exemplar from an exemplar repository. The exemplar repository stores a plurality of exemplars. Each of the exemplars includes an optimal recipe, an exemplar prompt, and an exemplar status data set. In some examples, the retrieved exemplar is chosen based on the similarity of the corresponding exemplar prompt and / or exemplar status data set to the user prompt and / or current status data set. In other examples, the retrieved exemplar is chosen based on the similarity of the optimal recipe to the active 2024PF80360

[0011] 2 lighting recipe of the lighting system. In even further examples, the retrieved exemplar is chosen based on a plurality of difference dimension metrics. The difference dimension metrics represent various modes in which the user prompt could vary from the exemplar prompts, such as a time / historical dimension, a context dimension, and / or a user dimension.

[0012] The system then determines deviation parameters by comparing the user prompt and / or the current status data set to the exemplar prompt and / or the exemplar status data set of the retrieved exemplar. The deviation parameters are then used to generate recipe adjustment limits to guide a large language model (LLM) in adjusting the retrieved optimal recipe to compensate for differences in prompting and status. Accordingly, the recipe adjustment limits prevent the LLM from generating an unsafe or otherwise undesirable lighting recipe by defining ranges of acceptable modifications to the optimal recipe.

[0013] The LLM then generates the lighting recipe based on the retrieved optimal recipe and the recipe adjustment limits. Accordingly, the lighting recipe is customized for the user prompt and / or the current status of the environment, the lighting system (and other connected aspects of the environment), and the user.

[0014] Generally, in one aspect, a method for generating a lighting recipe is provided. The method includes retrieving an exemplar from an exemplar repository comprising a plurality of exemplars. The exemplar is retrieved based on a user prompt and a current status data set. Each of the plurality of exemplars comprises an optimal recipe, an exemplar prompt, and an exemplar status data set.

[0015] The method further includes determining one or more deviation parameters based on a prompt deviation between the user prompt and the exemplar prompt of the retrieved exemplar and / or a status deviation between the current status data set and the exemplar status data set of the retrieved exemplar.

[0016] The method further includes determining one or more allowable recipe adjustment limits based on the one or more deviation parameters.

[0017] The method further includes generating, via an LLM, the lighting recipe based on the optimal recipe and the one or more allowable recipe adjustment limits.

[0018] According to an example, the method may further comprise receiving, via a user interface, the user prompt.

[0019] According to an example, the current status data set comprises an environmental data set, a configuration data set, and / or a user data set.

[0020] According to an example, the environmental data set is at least partially based on sensor data captured by one or more sensors corresponding to an environment. 2024PF80360

[0021] 3

[0022] According to an example, the configuration data set is based on an active lighting recipe data set and / or an automated lighting schedule data set.

[0023] According to an example, the configuration data set comprises a luminaire count.

[0024] According to an example, retrieving the exemplar from the exemplar repository further comprises: (1) determining a plurality of prompt similarity metrics based on the user prompt and the exemplar prompts associated with the plurality of exemplars, wherein each of the plurality of prompt similarity metrics corresponds to one of the plurality of exemplars; (2) determining a plurality of recipe similarity metrics based on an active lighting recipe data set and a plurality of optimal recipes of the plurality of exemplars, wherein each of the plurality of recipe similarity metrics corresponds to one of the plurality of exemplars; and (3) retrieving the exemplar based on the plurality of prompt similarity metrics and / or the plurality of recipe similarity metrics.

[0025] According to an example, the plurality of recipe similarity metrics for each of the plurality of exemplars are further based on comparing a user-accepted recipe and one of the plurality of optimal recipes.

[0026] According to an example, the determining the lighting recipe is further based on a deviation threshold 158 and at least one of an active lighting recipe or a user-accepted lighting recipe.

[0027] According to an example, retrieving the exemplar from the exemplar repository further comprises: (1) determining, via the LLM, a plurality of difference dimension metrics based on the plurality of the exemplar prompts and the user prompt; and (2) retrieving the exemplar based on the plurality of difference dimension metrics.

[0028] According to an example, determining the deviation parameters is further based on one or more difference dimension metrics of the plurality of the difference dimension metrics associated with the optimal recipe of the retrieved exemplar.

[0029] According to an example, the plurality of difference dimension metrics include a time / historical dimension, a context dimension, and / or a user dimension associated with each optimal recipe of the plurality of exemplars.

[0030] According to an example, the lighting recipe comprises one or more static lighting effects, one or more dynamic lighting effects, and / or one or more transitions between lighting effects.

[0031] According to an example, the retrieved exemplar is generated by merging, via the LLM, two or more optimal recipes into a homogenous recipe based on the user prompt. 2024PF80360

[0032] 4

[0033] Generally, in another aspect, a system for generating a lighting recipe is provided. The system comprises a controller configured to retrieve an exemplar from an exemplar repository comprising a plurality of exemplars. The exemplar is retrieved based on a user prompt and a current status data set. Each of the plurality of exemplars comprises an optimal recipe, an exemplar prompt, and an exemplar status data set.

[0034] The controller is further configured to determine one or more deviation parameters based on a prompt deviation between the user prompt and the exemplar prompt of the retrieved exemplar and / or a status deviation between the current status data set and the exemplar status data set of the retrieved exemplar.

[0035] The controller is further configured to determine one or more allowable recipe adjustment limits based on the one or more deviation parameters.

[0036] The controller is further configured to generate, via an LLM, the lighting recipe based on the optimal recipe and the one or more allowable recipe adjustment limits.

[0037] It should be appreciated that all combinations of the foregoing concepts and additional concepts discussed in greater detail below (provided such concepts are not mutually inconsistent) are contemplated as being part of the inventive subject matter disclosed herein. In particular, all combinations of claimed subject matter appearing at the end of this disclosure are contemplated as being part of the inventive subject matter disclosed herein. It should also be appreciated that terminology explicitly employed herein that also may appear in any disclosure incorporated by reference should be accorded a meaning most consistent with the particular concepts disclosed herein.

[0038] In various implementations, a processor or controller may be associated with one or more storage media (generically referred to herein as “memory,” e.g., volatile and non-volatile computer memory such as RAM, PROM, EPROM, EEPROM, floppy disks, compact disks, optical disks, magnetic tape, SSD, etc.). In some implementations, the storage media may be encoded with one or more programs that, when executed on one or more processors and / or controllers, perform at least some of the functions discussed herein. Various storage media may be fixed within a processor or controller or may be transportable, such that the one or more programs stored thereon can be loaded into a processor or controller so as to implement various aspects as discussed herein. The terms “program” or “computer program” are used herein in a generic sense to refer to any type of computer code (e.g., software or microcode) that can be employed to program one or more processors or controllers. 2024PF80360

[0039] 5

[0040] These and other aspects of the various embodiments will be apparent from and elucidated with reference to the embodiment s) described hereinafter.

[0041] BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In the drawings, like reference characters generally refer to the same parts throughout the different views. Also, the drawings are not necessarily to scale, emphasis instead generally being placed upon illustrating the principles of the various embodiments.

[0043] FIG. l is a functional block diagram of a controller for generating a lighting recipe via a large language model (LLM), in accordance with an example.

[0044] FIG. 2 is a variation of the functional block diagram of FIG. 1 showing functional aspects of the LLM in greater detail, in accordance with an example.

[0045] FIG. 3 is a functional block diagram of a system for generating a lighting recipe, in accordance with an example.

[0046] FIG. 4 is a functional block diagram of the exemplar selector shown in FIG. 2, in accordance with an example.

[0047] FIG. 5 is variation of the functional block diagram of the exemplar selector shown in FIG. 4, in accordance with an example.

[0048] FIG. 6 is a flowchart of a method for generating a lighting recipe, in accordance with an example.

[0049] DETAILED DESCRIPTION OF EMBODIMENTS

[0050] The present disclosure is generally directed to systems and methods for generating a lighting recipe for one or more luminaires of a lighting system installed in an environment. The system receives a user prompt regarding a desired lighting recipe (such as “Give me happy lighting”) and a current status data set corresponding to the status of the environment, the lighting system configuration, and the user. The system retrieves an exemplar from an exemplar repository storing a plurality of exemplars. Each of the exemplars includes an optimal recipe, an exemplar prompt, and an exemplar status data set. In some examples, the retrieved exemplar is chosen based on the similarity of the corresponding exemplar prompt and / or exemplar status data set to the user prompt and / or current status data set. The system then determines deviation parameters by comparing the user prompt and / or the current status data set to the exemplar prompt and / or the exemplar status data set of the retrieved exemplar. The deviation parameters are then used to generate recipe adjustment limits to guide a large language model (LLM) how to adjust the retrieved 2024PF80360

[0051] 6 optimal recipe to compensate for differences in prompting and status. The LLM then generates the lighting recipe based on the retrieved optimal recipe and the recipe adjustment limits. Accordingly, the lighting recipe is customized for the user prompt and / or the current status of the environment, the lighting system (and other connected aspects of the environment), and the user.

[0052] Turning now to the figures, FIG. l is a functional block diagram showing aspects of a controller 100 for generating a customized lighting recipe 102. Broadly, the controller 100 includes a memory for storing data and a processor for processing data. The controller 100 may also include a receiver, transmitter, and / or a transceiver for receiving and / or transmitting data. Further, while the controller 100 is depicted in FIG. 1 as a single component, in some examples, the various functions of the controller 100 may be distributed across more than one component. The lighting system includes a plurality of luminaires corresponding to an environment. The luminaires may be any type of lighting device such as lamps, light emitting diode (LED) strips, lasers, etc. The environment may be any possible type of location or area. In a preferred example, the environment may be a room, section, or portion of a home, such as a bedroom, living room, office, etc. While the lighting recipe 102 may be configured to control static and / or dynamic aspects of the luminaires (brightness, color temperature, illumination schedule, etc.), the lighting recipe 102 may also be configured to control one or more non-lighting aspects of the lighting system. For example, the lighting recipe 102 may cause the lighting system to render audio output and coordinate the lighting output to correspond to the audio output.

[0053] As shown in FIG. 1, the controller 100 generates the lighting recipe 102 for a specific environment using an LLM 101. Using the LLM 101 to generate the lighting recipe 102 may be considered to be an inference, rather than training, process. In some examples, the LLM 101 may be a multimodal LLM. A multimodal LLM may be able to process and understand data from several different types of input sources, such as text data, image data, audio data, etc. In some examples, the LLM 101 is a Generative Pre-trained Transformer 4 (GPT-4) model.

[0054] Three types of data are used by the LLM 101 to generate the lighting recipe 102: an exemplar 104, a user prompt 108, and a current status data set 110. A user prompt 108 is an input from a user instructing the LLM 101 to generate light. For example, the user prompt 108 could be the phrase “Give me happy lighting.” As will be demonstrated in FIG. 3, the user prompt 108 could be provided via any type of user interface 200 in wired or wireless communication with the controller 100. While a typical example user prompt 108 is 2024PF80360

[0055] 7 text-based, the user prompt 108 could be any type of data, including audio data, image data, video data, etc.

[0056] The current status data set 110 provides an array of data indicating the current state of the environment, the lighting system, and the user. For example, the current status data set 110 may include an environmental data set 126. The environmental data set 126 may convey information regarding the environment of the lighting system, such as time, date, location, temperature, humidity, occupancy, etc. As will be demonstrated with respect to FIG. 3, the environmental data 126 may be generated by one or more sensors 300.

[0057] The current status data set 110 may also include a configuration data set 128. The configuration data set 128 corresponds to the current configuration of the lighting system. The configuration data set 128 may include information such as luminaire count, lighting activation schedules, current brightness, current color temperature, etc. The configuration data set 128 may also include data regarding non-lighting systems within the environment, such as HVAC systems or audio systems. The configuration data set 128 may be generated based on information retrieved from the memory 125 of the controller 100. For example, the memory 125 may store an active lighting recipe data set 132 reflecting the lighting recipe currently being implemented in the environment and / or an automated lighting schedule data set 160 describing a schedule to activate and deactivate one or more luminaires of the lighting system.

[0058] The current status data set 110 may also include a user data set 130. The user data set 130 may include identification information of the user. This identification information could be retrieved from the memory 125 or derived from the sensors 300 using one or more identification techniques (facial identification, voice identification, etc.). The user data set 130 may also include a wide variety of data regarding the state of the user, such as a physical state or an emotional state. For example, the user data set 130 could indicate that the user is sleep deprived, returning from a hard day at work, alone in the environment, in the room with their spouse, or about to leave the environment. The user data set 130 could also indicate predicted future user states, such as the user being about to return from work. As with the identification information, information regarding the physical or emotional state of the user may be derived from information stored in memory 125 or derived from the sensors 300.

[0059] The exemplar 104 includes an optimal recipe 106, an exemplar prompt 112, and an exemplar status data set 114. The optimal recipe 106 is an example lighting recipe which has been previously implemented. As with the previously described lighting recipe 2024PF80360

[0060] 8

[0061] 102, the optimal recipe 106 may define a wide array of lighting and non-lighting parameters, such as brightness, color temperature, illumination schedule, etc. This example lighting recipe may been implemented upon the lighting system currently being configured or on a different lighting system in a different environment. Each optimal recipe 106 is associated with an exemplar prompt 112 and an exemplar status data set 114 indicating the prompt and statuses corresponding to the optimal recipe 106. The optimal recipes 106 of the exemplars 104 may be handcrafted by users or technicians or synthetic recipes generated via computer processing.

[0062] The exemplar 104 is stored with a plurality of other exemplars 104 in an exemplar repository 50. The exemplar repository 50 may be embodied as a vector database. In some examples, the exemplar repository 50 may be created, at least in part, by the LLM 101. As will be demonstrated below, the LLM 101 retrieves the exemplar 104 from the exemplar repository 50 based on the user prompt 108 and the current status data set 110. Thus, if an exemplar 104 has an exemplar prompt 112 and an exemplar status data set 114 matching the provided user prompt 108 and current status data set 110, the optimal recipe 106 of the exemplar 104 should provide lighting effects desired by the user. However, in many cases, the exemplar repository 50 may not contain an exemplar 104 with an exemplar prompt 112 and an exemplar status data set 114 exactly matching the user prompt 108 and the current status data set 110. Accordingly, the LLM 101 is typically configured to (1) choose the exemplar 104 with an exemplar prompt 112 and an exemplar status data set 114 closest to the user prompt 108 and the current status data set 110 and then (2) adjust the associated optimal recipe 106 based on the differences between the user prompt 108 and the exemplar prompt 112 as well as the current status data set 110 and the exemplar status data set 114.

[0063] As will be described below, the adjustment of the optimal recipe 106 is controlled or limited by allowable recipe adjustment limits 122. The recipe adjustment limits 122 may be based on differences between the exemplar prompt 112 and the user prompt 108 and / or differences between the exemplar status data set 114 and the current status data set 110 to allow the LLM to adjust the optimal recipe 106 within ranges defined by the recipe adjustment limits 122. For example, the recipe adjustment limits 122 may prevent the LLM 101 from increasing the brightness of the optimal recipe 106 too high for either the particular user or the environment.

[0064] FIG. 2 conceptually illustrates the flow of data to the various functions of the LLM 101 in greater detail. In particular, the LLM 101 is shown as implementing an exemplar 2024PF80360

[0065] 9 selector 103, a deviation calculator 105, a limit translator 107, and a recipe adjustor 109. The exemplar selector 103 attempts to find an exemplar 104 which best corresponds to the current situation within the environment. The exemplar selector 103 selects or chooses an exemplar 104 from the exemplar repository 50 shown in FIG. 1 based on the user prompt 108 and the current status data set 110. In particular, the exemplar selector 103 may compare the provided user prompt 108 to the exemplar prompts 112 associated with each of the exemplars 104 and select the exemplars 104 having the exemplar prompt 112 closest to the user prompt 108. Further, the exemplar selector 103 may compare the provided current status data set 110 to the exemplar status data set 114 associated with each of the exemplars 104 and select the exemplar 104 having the current status data set 114 closest to the current status data set 110. In further examples, the exemplar 104 may be chosen based on a combination of both of these comparisons. These comparisons are described in more detail with regard to FIGS. 4 and 5.

[0066] With an exemplar 104 selected, the deviation calculator 105 determines one or more deviation parameters 116. These deviation parameters 116 include (1) a prompt deviation 118 quantifying the difference between the user prompt 108 and the exemplar prompt 112 associated with the selected exemplar 104 and (2) a status deviation 120 quantifying the difference between the current status data set 110 and the exemplar status data set 112 associated with the selected exemplar 104. For example, the prompt deviation 118 could quantify the difference between the user prompt 108 of “Give me happy lighting” and the exemplar prompt 112 of “I want to feel happy again.” The exemplar prompt 112 indicates that the previous user indicated that they were previously happy, then became unhappy, and then wished to become happy once again. In both cases, the resulting lighting effect end state may be the same, lighting with a “happy” ambience. However, the deviation calculator 105 identifies that the exemplar prompt 112 includes the word “again” and quantifies this difference as part of the prompt deviation 118. Similarly, the status deviation 120 may quantify similar differences in various status characteristics, such as day of the year / month / week, time of day, lighting system luminaire count, characteristics and locations, emotional status of the user, etc. For example, the status deviation 120 may quantify that the lighting system corresponding to the exemplar 104 used five luminaires (based on the luminaire count of the exemplar status data set 114, while the lighting system to receive the lighting recipe 102 only has three luminaires (based on the luminaire count of the configuration data 128). In some cases, this difference in luminaire count could occur due to one or more luminaires being unplugged. 2024PF80360

[0067] 10

[0068] The deviation parameters 116 are then provided to the limit translator 107. The limit translator 107 converts the deviation parameters 116 (including the prompt deviation 118 and / or the status deviation 120) into allowable recipe adjustment limits 122 to guide the finetuning of the optimal recipe 106 of the selected exemplar 104. For example, the prompt deviation 118 may indicate that the current user prompt 108 includes the word “again,” while the exemplar prompt 112 does not. To account for this difference, the recipe adjustment limits 122 may allow for a more gradual transition from an initial lighting state to a final lighting state. Further, the status deviation 120 may indicate that the current lighting system has less luminaires than the system which implemented the optimal recipe 106. Accordingly, the recipe adjustment limits 122 may allow for an increase to the brightness of each of the three luminaires to generate total illumination similar to the five luminaires of the exemplar system.

[0069] In some examples, the deviation parameters 116 may be compared to one or more thresholds to determine if an adjustment is necessary. For example, the prompt deviation 118 may be compared to a prompt threshold, while the status deviation 120 may be compared to a deviation threshold. If the prompt deviation 118 and the status deviation 120 are both below their corresponding thresholds, the optimal recipe 106 does not need to be adjusted.

[0070] The recipe adjustment limits 122 are then provided to the recipe adjustor 109. The recipe adjustor 109 modifies the optimal recipe 106 within the bounds defined by the recipe adjustment limits 122 to create a finetuned, customized lighting recipe 102. The recipe adjustment limits 122 may provide one or more limitations on the degree of change or final values of the lighting recipe 102. For example, while the LLM 101 may be inclined to increase the brightness of the luminaires to generate a desired lighting effect, the allowable recipe adjustment limits 122 may indicate that the brightness increase should be limited due to one or more factors (such as luminaire specification, user brightness preferences, etc.). In another example, the recipe adjustment limits 122 may indicate that while brightness may be increased during evening hours, a warm white correlated color temperature (CCT) of a lighting scene must remain constant during the same evening hours. In this way, the recipe adjustment limits 122 may represent the allowable extent to which the LLM 101 may modify the optimal recipe 106 of the selected exemplar 104.

[0071] In further examples, the recipe adjustor 109 may further adjust the optimal recipe 106 according to one or more adjustment exemplars 154. Rather than representing previously implemented lighting recipes, these adjustment exemplars 154 teach the recipe 2024PF80360

[0072] 11 adjustor 109 how to modify, finetune, and / or customize the optimal recipe 106 based on additional factors. These adjustment exemplars 154 may be retrieved based on the user prompt 108, the current status data set 110, or additional information.

[0073] For instance, a selected exemplar 104 may closely correspond to the user prompt 108 and the current status data set 110. However, despite this close correspondence, further finetuning may be required due to additional information. For example, based on additional information, the optimal recipe 106 may be evaluated as being “hyper-active” for the user due to quick changes in dynamic and static lighting effects. Accordingly, an adjustment exemplar 154 may be used to instruct the recipe adjustor 109 to “calm” the optimal recipe 106 of the exemplar 104 by reducing the variability in the time domain and / or by increasing uniformity in the spatial domain.

[0074] In another example, an adjustment exemplar 154 may be used to effectuate a balance between a desired lighting effect described in the user prompt 108 and potentially negative health impacts associated with the lighting effects. In this example, the exemplar selector 103 may choose an exemplar 104 having an optimal recipe 106 defining a stroboscopic effect. The adjustment exemplar 154 may show the recipe adjustor 109 how to reduce the potentially dangerous aspects of the stroboscopic effect while maintaining an immersive user experience.

[0075] In a further example, an adjustment exemplar 154 may be used to teach the recipe adjustor 109 when to consider the active lighting recipe 132 to limit the adjustment of the optimal recipe 106 of the exemplar 104. For example, if the user prompt 108 is related to improving user comfort, the adjustment exemplar 154 instructs the recipe adjustor 109 to create a finetuned lighting recipe 102 relatively close to the active lighting recipe 132. In other examples, the adjustment exemplar 154 may enable the recipe adjustor 109 to perform radical changes to the active lighting recipe 132 in certain circumstances, such as to indicate the occurrence of a health or safety situation requiring emergency assistance.

[0076] In even further examples, an adjustment exemplar 154 may be used to limit the adjustment of the optimal recipe 106 to a subset of parameters. For example, the adjustment exemplar 154 could instruct the recipe adjustor 109 to only adjust a color palette selection and a level of dynamic effects. Other parameters, such as brightness or color temperature must remain the same.

[0077] In some examples, the adjustment of the optimal recipe 106 may be further based on a deviation threshold 158. In some circumstances, a perceptible change from the current, active lighting recipe 132 is desired. Accordingly, a deviation threshold 158 may be 2024PF80360

[0078] 12 used to ensure that the difference between the output lighting recipe 102 and the active lighting recipe exceeds the deviation threshold 158. Alternatively, in other circumstances, a perceptible change is not desired. In these cases, the deviation threshold 158 may be used to ensure that the difference between the output lighting recipe 102 and the active lighting recipe does not exceed the deviation threshold 158. In further circumstances, the output lighting recipe 102 may be compared to a user-accepted recipe 156. The user-accepted recipe 156 may have been previously created and / or fine-tuned based on user input, or the user- accepted recipe 156 may have been synthetically created but manually approved by the current user. Thus, the deviation threshold 158 may be used control how much or how little the output lighting recipe 102 deviates from the user-accepted lighting recipe 156.

[0079] In some examples, the exemplar selector 103 may retrieve multiple exemplars 104 from the exemplar repository. For example, a first exemplar 104a could include a first optimal recipe 106a defining an entertainment lighting scene for a left side of a room, while a second exemplar 104b could include a second optimal recipe 106b defining an entertainment lighting scene for a right side of the same room. By comparing the user prompt 108 to the exemplar prompts 112, the exemplar selector 103 may determine that more than one exemplar 104 is required. Further, in this example, a third exemplar 104c may be selected which illustrates how to merge the left side optimal recipe 106a with the right-side optimal recipe 106b into a homogenous, whole room lighting recipe. Thus, in this example, the exemplar 104 outputted by the exemplar selector 103 is a homogenous merger of the left side exemplar 104a and the right-side exemplar 104b. The optimal recipe 106 of the merged exemplar 104 may then be finetuned according to the user prompt 108 and current status data set 110 as described above.

[0080] In a further example, rather than merging exemplars 104a, 104b corresponding to different areas of a physical space, the exemplar selector 103 may merge two or more exemplars 104 in the time domain. These merged exemplars 104 may each contain dynamic lighting effects to create a unique, time-varying lighting recipe to be further finetuned according to the user prompt 108 and current status data set 110. In one example, the optimal recipe 106a of the first exemplar 104a may define a “wake-up” scene, and the optimal recipe 106b of the second exemplar 104b may define an “energize” scene. Thus, the merged exemplar 104 may start with the wake-up scene, and then seamlessly transition into the energize scene.

[0081] As previously noted, the finetuned lighting recipe 102 may also generate nonlighting effects, such as audio outputs. In one example, an exemplar 104 in the exemplar 2024PF80360

[0082] 13 repository 50 may have an exemplar prompt 112 of “Start playing music and sync my light to it.” Accordingly, the optimal recipe 106 of the exemplar 104 plays audio via a streaming service and automatically selects settings for color palette and dynamic effects based on the current song being played. However, the user prompt 108 received by the LLM 101 may be slightly different than the exemplar prompt 112. For example, the user prompt 108 may be “Cheer me up, start playing music and sync my light to it.” By noting the difference between the user prompt 108 and the exemplar prompt 112, the optimal recipe 106 may be adjusted appropriately. While the lighting effects will still be synchronized with the music, the color palette may be modified to focus on “happy” colors, and the dynamic effects may be similarly modified to fit with “happy” music.

[0083] In some examples, the functions of the deviation calculator 105, the limit translator 107, and the recipe adjustor 109 may occur outside of the LLM 101. In these examples, the LLM 101 is used to retrieve an exemplar 104, and then provides the optimal recipe 106 (and possibly other aspects of the exemplar 104) to an external system or device, such as the lighting controller 400, for finetuning. For example, the LLM 101 may be used to retrieve the exemplar 104 and then determine the one or more deviation parameters 116. In this example, the calculated deviation parameters 116 are quantifiable, in that the deviation parameters 116 may be provided to a non-LLM function to update the optimal recipe 106. Thus, rather than providing the deviation parameters 116 to the limit translator 107, the LLM 101 retrieves or calls a non-LLM function and inputs the deviation parameters 116 to the function. In a particular example, the user prompt 108 may request a “dim sunset,” and the selected exemplar 104 (based on a vector similarity search of the exemplar repository 50) is simply “sunset.” The LLM 101 produces deviation parameters 116 identifying the “dim” aspect of the user prompt 108 as a deviation from the optimal recipe 106 of the selected exemplar 104. The LLM 101 then reviews a pool of potential functions (such as brightness, colorfulness, color temperature, etc.) and identifies “brightness” as the function appropriate to address the “dim” deviation. Based on the “dim” deviation, the function reduces (such as by 50%) the brightness of the optimal recipe 106 to produce a dimmed effect. The modified optimal recipe is then provided to the lighting system as the lighting recipe 102. Further, the modified optimal recipe could be stored as a new exemplar 104 in the exemplar repository 50. In other variations, if no appropriate functions are available, the LLM 101 may instead generate a new lighting recipe 102 as described above.

[0084] FIG. 3 is a further example of a functional block diagram of a system 10 for generating a lighting recipe 102. In particular, the system 10 of FIG. 3 shows the inputs into 2024PF80360

[0085] 14 the LLM 101 in greater detail. For example, FIG. 3 shows the user prompt 108 being provided to the LLM 101 by a user interface 200. The user interface 200 may be any type of user interface, such as a keyboard, a touch screen, microphone, or any other interface capable of receiving the user prompt 108. In some examples, the user interface 200 is integrated into the previously described controller 100. In other examples, the user interface 200 may be a discrete component that communicates with the LLM 101 via one or more wired and / or wireless connections.

[0086] FIG. 3 further shows a current status aggregator 111 configured to generate the current status data set 110. As previously described, the current status data set 110 may include the environmental data set 126 corresponding to the environment of the lighting system, the configuration data set 128 corresponding to the lighting system, and the user data set 130 corresponding to the user. In some examples, the current status data set 110 is at least partially based on sensor data 302 provided by one or more sensors 300. The sensors 300 may include temperature sensors, humidity sensors, occupancy sensors, etc. In some examples, the sensors 300 are integrated into the previously described controller 100. In other examples, the sensors 300 may be discrete components that communicate with the LLM 101 via one or more wired and / or wireless connections.

[0087] Further, the current status data set 110 may be based on status data 150 stored in the memory 125. The status data 150 may include an active lighting recipe 132 currently being implemented and / or an automated lighting schedule data set 160 describing a schedule of activate and deactivating one or more luminaires of the lighting system. The automated lighting schedule data set 160 may also include timing information regarding other lighting settings, such as when to implement certain CCT settings. While the memory 125 may be considered a component of the controller 100, in other examples, the memory 125 could be an external memory component that communicates with the LLM 101 via one or more wired and / or wireless connections.

[0088] FIG. 3 further shows a lighting controller 400. The lighting controller 400 is configured to implement the finetuned lighting recipe 102 using one or more luminaires of a lighting system in an environment, such as a room of a home. The lighting recipe 102 may be defined by one or more static lighting effects 146 (brightness, color temperature, etc.) and one or more dynamic lighting effects 148 (change in brightness over time, etc.).

[0089] FIG. 4 illustrates the conceptual exemplar selector 103 of FIG. 2 in greater detail. In the example of FIG. 4, the exemplar 104 is selected based on a combination of the user prompt 108 and a lighting recipe, such as an active lighting recipe 132 or a user- 2024PF80360

[0090] 15 accepted recipe 156. As shown in FIG. 4, a prompt comparator 115 compares the user prompt 108 to the exemplar prompts 112 to determine a plurality of prompt similarity metrics 134. Each prompt similarity metric 134 represents how close an exemplar prompt 112 is to the user prompt 108. In some examples, the prompt similarity metrics 134 includes a probability value representing the likelihood that the user prompt 108 is equivalent to an exemplar prompt 112.

[0091] Further, a recipe comparator 117 compares the active lighting recipe 132 (which is an aspect of the current status data set 110) to the optimal recipes 106 of the stored exemplars 104 to determine a plurality of recipe similarity metrics 136. Each recipe similarity metric 136 represents how close an optimal recipe 106 is to the active lighting recipe 132.

[0092] Alternatively, the recipe comparator 117 may instead compare the optimal recipes 106 to a user-accepted recipe 156. The user-accepted recipe 156 may have been previously created and / or fine-tuned based on user input, or the user-accepted recipe 156 may have been synthetically created but manually approved by the current user. Thus, the recipe similarity metric 136 for each exemplar 104 may represent the similarity of the user-accepted recipe 156 to the optimal recipe 106 of the same exemplar 104. In other examples, the recipe similarity metric 136 may synthesize the comparison of the optimal recipe 106 to both the active lighting recipe 132 and the user-accepted recipe 156.

[0093] An exemplar identifier 119 then uses the prompt similarity metrics 134 and the recipe similarity metrics 136 to select an exemplar 104. In one embodiment, the exemplar identifier 119 may choose the exemplar 104 corresponding to the highest prompt similarity metric 134. Further, the fact that the user has entered a user prompt 108 typically indicates that the user wants a significant change to occur to the lighting system. Therefore, the optimal recipe 106 of the selected exemplar 104 should not be too similar to the active lighting recipe 132. Accordingly, the recipe similarity metric 136 corresponding to the exemplar 104 having the highest prompt similarity metric 134 may then be compared to a recipe similarity threshold 152. If the selected exemplar 104 is sufficiently different, the selected exemplar 104 is outputted by the exemplar selector 103 for further processing. If the selected exemplar 104 is too similar, optimal recipes 106 of exemplars 104 having the next highest prompt similarity metric 134 may be chosen and evaluated for similarity to the active lighting recipe 132. Other variations of this processing method may be used to select an exemplar 104 associated with an exemplar prompt 112 similar to the user input 108, while the associated optimal recipe 106 is sufficiently different from the active lighting recipe 132 currently being implemented in the environment. In even further variations, the user may 2024PF80360

[0094] 16 only desire for small, relatively insignificant changes to the active lighting recipe 132. Accordingly, the recipe similarity threshold 152 may be used to select an exemplar 104 having a recipe similarity metric 136 below the recipe similarity threshold 152.

[0095] FIG. 5 illustrates another embodiment of the conceptual exemplar selector 103. In FIG. 5, a difference dimension calculator 121 determines a plurality of difference dimension metrics 134 by comparing the user prompt 108 to each of the exemplar prompts 112. The difference dimension metrics 134 may include a number of different dimensions defining how the user prompt 108 may differ from the exemplar prompts 112. For example, a time / historical dimension 140 represents the differences in the time domain, such as if the user prompt 108 refers to the past (“I feel happy again”) while the exemplar prompt 112 does not (“I feel happy”). Further, a context dimension 142 represents differences in the situation (such as the current configuration of the lighting system and the greater environment) described in the user prompt 108 and the exemplar prompt 112. Additionally, a user dimension 144 represents differences between the user associated with the user prompt 108 and the user associated with the exemplar prompt 112 as evidenced within the prompts themselves. These difference dimension metrics 134 are then provided to the exemplar identifier 119 and used to select an exemplar 104 to output. The exemplar 104 may be selected based on one of or a combination of the different difference dimension metrics 134. Further, in making this selection, some types of difference dimension metrics 134 may be weighted more than others.

[0096] In some examples, the difference dimension metrics 134 may also be passed onto the limit translator 107 shown in FIG. 2. The limit translator 107 may then use the difference dimensions 134 to calculate the allowable recipe adjustment limits 122 used to finetune the optimal recipe 106 into the customized lighting recipe 102.

[0097] In some examples, instead of exemplars 104, the exemplar repository 50 may also store pre-cursor images or pre-cursor videos. One or more the images and / or videos may be retrieved and modified based on the user prompt 108 and current status data set 110. The resulting images and / or videos are then provided to a recipe converter to transform the images and / or videos into the lighting recipe 102. For example, a user prompt 108 may cause the exemplar selector 103 to retrieve a first image of a tropical sunset and a second image of a tornado. These images may be merged to form a single image of a tornado during a tropical sunset. This merged image is then converted into the customized lighting recipe 102.

[0098] FIG. 6 illustrates a method 900 for generating a lighting recipe. With respect to FIGS. 1-6, the method 900 includes, in step 902, retrieving an exemplar 104 from an 2024PF80360

[0099] 17 exemplar repository 50 comprising a plurality of exemplars 104, wherein the exemplar 104 is retrieved based on a user prompt 108 and a current status data set 110, wherein each of the plurality of exemplars 104 comprises an optimal recipe 106, an exemplar prompt 112 and an exemplar status data set 114.

[0100] The method 900 further includes, in step 904, determining one or more deviation parameters 116 based on a prompt deviation 118 between the user prompt 108 and the exemplar prompt 110 associated with the retrieved exemplar 104 and / or a status deviation 120 between the current status data set 110 and the exemplar status data set 114 of the retrieved exemplar 104.

[0101] The method 900 further includes, in step 906 determining one or more allowable recipe adjustment limits 122 based on the one or more deviation parameters 116.

[0102] The method 900 further includes, in step 908, generating, via an LLM 101, the lighting recipe 102 based on the optimal recipe 106 and the one or more allowable recipe adjustment limits 122.

[0103] According to an example, the method 900 may further include, in optional step 910, receiving, via a user interface 200, the user prompt 108.

[0104] According to an example, the method 900 may further include, in optional step 912, implementing the lighting recipe 102 via one or more luminaires of a lighting system.

[0105] All definitions, as defined and used herein, should be understood to control over dictionary definitions, definitions in documents incorporated by reference, and / or ordinary meanings of the defined terms.

[0106] The indefinite articles “a” and “an,” as used herein in the specification and in the claims, unless clearly indicated to the contrary, should be understood to mean “at least one.”

[0107] The phrase “and / or,” as used herein in the specification and in the claims, should be understood to mean “either or both” of the elements so conjoined, i.e., elements that are conjunctively present in some cases and disjunctively present in other cases. Multiple elements listed with “and / or” should be construed in the same fashion, i.e., “one or more” of the elements so conjoined. Other elements may optionally be present other than the elements specifically identified by the “and / or” clause, whether related or unrelated to those elements specifically identified.

[0108] As used herein in the specification and in the claims, “or” should be understood to have the same meaning as “and / or” as defined above. For example, when separating items in a list, “or” or “and / or” shall be interpreted as being inclusive, i.e., the 2024PF80360

[0109] 18 inclusion of at least one, but also including more than one, of a number or list of elements, and, optionally, additional unlisted items. Only terms clearly indicated to the contrary, such as “only one of’ or “exactly one of,” or, when used in the claims, “consisting of,” will refer to the inclusion of exactly one element of a number or list of elements. In general, the term “or” as used herein shall only be interpreted as indicating exclusive alternatives (i.e. “one or the other but not both”) when preceded by terms of exclusivity, such as “either,” “one of,” “only one of,” or “exactly one of.”

[0110] As used herein in the specification and in the claims, the phrase “at least one,” in reference to a list of one or more elements, should be understood to mean at least one element selected from any one or more of the elements in the list of elements, but not necessarily including at least one of each and every element specifically listed within the list of elements and not excluding any combinations of elements in the list of elements. This definition also allows that elements may optionally be present other than the elements specifically identified within the list of elements to which the phrase “at least one” refers, whether related or unrelated to those elements specifically identified.

[0111] It should also be understood that, unless clearly indicated to the contrary, in any methods claimed herein that include more than one step or act, the order of the steps or acts of the method is not necessarily limited to the order in which the steps or acts of the method are recited.

[0112] In the claims, as well as in the specification above, all transitional phrases such as “comprising,” “including,” “carrying,” “having,” “containing,” “involving,” “holding,” “composed of,” and the like are to be understood to be open-ended, i.e., to mean including but not limited to. Only the transitional phrases “consisting of’ and “consisting essentially of’ shall be closed or semi-closed transitional phrases, respectively.

[0113] The above-described examples of the described subject matter can be implemented in any of numerous ways. For example, some aspects may be implemented using hardware, software, or a combination thereof. When any aspect is implemented at least in part in software, the software code can be executed on any suitable processor or collection of processors, whether provided in a single device or computer or distributed among multiple devices / computers.

[0114] The present disclosure may be implemented as a system, a method, and / or a computer program product at any possible technical detail level of integration. The computer program product may include a computer readable storage medium (or media) having 2024PF80360

[0115] 19 computer readable program instructions thereon for causing a processor to carry out aspects of the present disclosure.

[0116] The computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium may be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. A non- exhaustive list of more specific examples of the computer readable storage medium includes the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or raised structures in a groove having instructions recorded thereon, and any suitable combination of the foregoing. A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.

[0117] Computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network may comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing / processing device.

[0118] Computer readable program instructions for carrying out operations of the present disclosure may be assembler instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, configuration data for integrated circuitry, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++, or the like, and 2024PF80360

[0119] 20 procedural programming languages, such as the “C” programming language or similar programming languages. The computer readable program instructions may execute entirely on the user’s computer, partly on the user's computer, as a stand-alone software package, partly on the user’s computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider). In some examples, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate arrays (FPGA), or programmable logic arrays (PLA) may execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present disclosure.

[0120] Aspects of the present disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to examples of the disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer readable program instructions.

[0121] The computer readable program instructions may be provided to a processor of a, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks. These computer readable program instructions may also be stored in a computer readable storage medium that can direct a computer, a programmable data processing apparatus, and / or other devices to function in a particular manner, such that the computer readable storage medium having instructions stored therein comprises an article of manufacture including instructions which implement aspects of the function / act specified in the flowchart and / or block diagram or blocks.

[0122] The computer readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process, such that the instructions which execute 2024PF80360

[0123] 21 on the computer, other programmable apparatus, or other device implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0124] The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various examples of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the blocks may occur out of the order noted in the Figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flowchart illustration, and combinations of blocks in the block diagrams and / or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts or carry out combinations of special purpose hardware and computer instructions.

[0125] Other implementations are within the scope of the following claims and other claims to which the applicant may be entitled.

[0126] While various examples have been described and illustrated herein, those of ordinary skill in the art will readily envision a variety of other means and / or structures for performing the function and / or obtaining the results and / or one or more of the advantages described herein, and each of such variations and / or modifications is deemed to be within the scope of the examples described herein. More generally, those skilled in the art will readily appreciate that all parameters, dimensions, materials, and configurations described herein are meant to be exemplary and that the actual parameters, dimensions, materials, and / or configurations will depend upon the specific application or applications for which the teachings is / are used. Those skilled in the art will recognize, or be able to ascertain using no more than routine experimentation, many equivalents to the specific examples described herein. It is, therefore, to be understood that the foregoing examples are presented by way of example only and that, within the scope of the appended claims and equivalents thereto, examples may be practiced otherwise than as specifically described and claimed. Examples of the present disclosure are directed to each individual feature, system, article, material, kit, and / or method described herein. In addition, any combination of two or more such features, systems, articles, materials, kits, and / or methods, if such features, systems, articles, materials, 2024PF80360

[0127] 22 kits, and / or methods are not mutually inconsistent, is included within the scope of the present disclosure.

Claims

2024PF8036023CLAIMS1. A method (900) for generating a lighting recipe, comprising: retrieving (902) an exemplar from an exemplar repository comprising a plurality of exemplars, wherein the exemplar is retrieved based on a user prompt and a current status data set, wherein each of the plurality of exemplars comprises an optimal recipe, an exemplar prompt, and an exemplar status data set; determining (904) one or more deviation parameters based on a prompt deviation between the user prompt and the exemplar prompt of the retrieved exemplar quantifying a difference between the user prompt and the exemplar prompt and / or a status deviation between the current status data set and the exemplar status data set of the retrieved exemplar quantifying a difference between the current status data set and the exemplar status data set; determining (906) one or more allowable recipe adjustment limits based on the one or more deviation parameters; and generating, (908) via a large language model (LLM), the lighting recipe based on the optimal recipe and the one or more allowable recipe adjustment limits.

2. The method (900) of claim 1, further comprising receiving, via a user interface, the user prompt.

3. The method (900) of claim 1, wherein the current status data set comprises an environmental data set, a configuration data set, and / or a user data set.

4. The method (900) of claim 3, wherein the environmental data set is at least partially based on sensor data captured by one or more sensors corresponding to an environment.

5. The method (900) of claim 3, wherein the configuration data set is based on an active lighting recipe data set and / or an automated lighting schedule data set.2024PF80360246. The method (900) of claim 3, wherein the configuration data set comprises a luminaire count.

7. The method (900) of claim 1, wherein retrieving the exemplar from the exemplar repository further comprises: determining a plurality of prompt similarity metrics based on the user prompt and the exemplar prompts associated with the plurality of exemplars, wherein each of the plurality of prompt similarity metrics corresponds to one of the plurality of exemplars; determining a plurality of recipe similarity metrics based on an active lighting recipe data set and a plurality of optimal recipes of the plurality of exemplars, wherein each of the plurality of recipe similarity metrics corresponds to one of the plurality of exemplars; and retrieving the exemplar based on the plurality of prompt similarity metrics and / or the plurality of recipe similarity metrics.

8. The method (900) of claim 7, wherein the plurality of recipe similarity metrics for each of the plurality of exemplars are further based on comparing a user-accepted recipe and one of the plurality of optimal recipes.

9. The method (900) of claim 1, wherein the determining the lighting recipe is further based on a deviation threshold 158 and at least one of an active lighting recipe or a user-accepted lighting recipe.

10. The method (900) of claim 1, wherein retrieving the exemplar from the exemplar repository further comprises: determining, via the LLM, a plurality of difference dimension metrics based on the plurality of the exemplar prompts and the user prompt; retrieving the exemplar based on the plurality of difference dimension metrics.

11. The method (900) of claim 10, wherein determining the deviation parameters is further based on one or more difference dimension metrics of the plurality of the difference dimension metrics associated with the optimal recipe of the retrieved exemplar.2024PF803602512. The method (900) of claim 10, wherein the plurality of difference dimension metrics include a time / historical dimension, a context dimension, and / or a user dimension associated with each optimal recipe of the plurality of exemplars.

13. The method (900) of claim 1, wherein the lighting recipe comprises one or more static lighting effects, one or more dynamic lighting effects, and / or one or more transitions between lighting effects.

14. The method (900) of claim 1, wherein the retrieved exemplar is generated by merging, via the LLM, two or more optimal recipes into a homogenous recipe based on the user prompt.

15. A system (10) for generating a lighting recipe (102) comprising a controller (100) configured to: retrieve an exemplar (104) from an exemplar repository (50) comprising a plurality of exemplars (104), wherein the exemplar (104) is retrieved based on a user prompt (108) and a current status data set (110), wherein each of the plurality of exemplars (104) comprises an optimal recipe (106), an exemplar prompt (112), and an exemplar status data set (H4); determine (904) one or more deviation parameters (116) based on a prompt deviation (118) between the user prompt (108) and the exemplar prompt (112) of the retrieved exemplar (104) quantifying a difference between the user prompt and the exemplar prompt and / or a status deviation (120) between the current status data set (110) and the exemplar status data set (114) of the retrieved exemplar (104) quantifying a difference between the current status data set and the exemplar status data set; determining (906) one or more allowable recipe adjustment limits (122) based on the one or more deviation parameters (116); and generating, (908) via a large language model (LLM), the lighting recipe (102) based on the optimal recipe (106) and the one or more allowable recipe adjustment limits (122).

Images